Full text of "Cardona & Zide (eds.) - Festschrift for Henry Hoenigswald (1987)"

See other formats

eters

a
E
SCH
te
e
De
DO
di
d
i
qi 4
a Wa
Go ee
do d Gei 4
BR dnd
a i
27
o
Go
VU 4,
d
ii
i
200 n DODO
EE
N,
d x
1
A
d à » y Ke

5 NN NON

Jj
Vy AIDE ORE
ent 2)
GENEE
KEE
ON? SC
ER SS
EE G i o

n
v Die) "i D

A
ERS

ht bb

Seabee ee ee ee
CN RR
dan TAI E
Ke GE oi di
IS PTS
lg

WEG,
E
SE
E
à eie e t
aH A MEUM AT
ER,

ANNEES
ias lut nt
di deu!

de,
00000
d NN

Ste
Seen Eat

We
G DEENEN
DEINEN #

4 eo Je

VP RA
ANH IH
und
RT

I $

EEE TEE,

> 5
ERE

ai
Sn

=
ES

Ge
sui
da

=»

ne
na

#2
RUE

a es

"|
OUO
ES :

je”
SN

SEA

EEN
GA

WË
DOO
eub als

TI
pinto

STE

Sasa

a
ani

EC
T
SE
M
rn
ra

rana
eis
ais

A
ee
ETERO

pnt
e e e Ae

n:
=

E
EE

È a
a

URI + e OX

tt Re s

aere en Gu ON
SE

GAN?

nu See
Wd dée ELBE

Wees ry BE ^ "n asd Ge

nee

E
EE Di
PUPENE

5
Ua n

T ME UNE

SIRIA
KANN
RR)
COCO

d Kee

Weder

Weeer
dt
3

iet

A
ERBE
"ret.

d
inni

i
epee Eet

Weist dree
snes

nun

t
at

4
A

j^
GE

Grey i

GC
=
TR]

T
Sa

LS
SS

a ete

MIS
<<

MAE,

na
SS
HUE

DH) d

BEREITEN

IRRE
Gd

RR
MIRA UH A A
ORK dee

DNA a

SEA
i

"7
Veedel
SH

DODUDO
RER
PS,
(EC
Ae)

a
WI

dut.

M VI

Va Fa

DO
ab ht

4
ETE

VA RUE
DEN AI. A

Festschrift for Henry Hoenigswald

Ars Linguistica 15

Commentationes analyticae et criticae

Editor: Werner Winter

Festschrift for
Henry Hoenigswald

On the Occasion
of his Seventieth Birthday

Edited by

George Cardona
and
Norman H. Zide

gnV Gunter Narr Verlag Tübingen

Deutsche Bibliothek Cataloguing in Publication Data

Festschrift for Henry Hoenigswald : on the occasion of his 70. birthday /
ed. by George Cardona and Norman H. Zide. — Tübingen : Narr, 1987.
(Ars linguistica ; 15)
ISBN 3-87808—365—3

NE: Cardona, George [Hrsg.]; Hoenigswald, Henry: Festschrift; GT

All rights, including the rights of publication, distribution and sales, as well as the right
to translation, are reserved. No part of this work covered by the copyrights hereon may
be reproduced or copied in any form or by any means — graphic, electronic or
mechanical including photocopying, recording, taping, or information and retrieval
systems — without written permission of the publisher.

Printed in Germany

ISBN 3-87808—365-3

FOREWORD

We are happy to pay homage to Henry Hoenigswald with this volume of studies by col-
leagues and friends, several of whom have had the privilege of studying with him. It is
fitting that scholars from all over the world should join in expressing their esteem for
a colleague whose many varied and fundamental contributions to historical linguistics
and the history of linguistics are so widely known and appreciated:

vidvän sarvatra pujyate

Tabula gratulatoria

Francisco R. Adrados, Madrid
William Sidney Allen, Cambridge
Yoél L. Arbeitman, New York
Harold W. Bailey, Cambridge
Philip Baldi, Philadelphia
Johannes Bechert, Bremen
Robert S.P. Beekes, Leiden
Allan R. Bomhard, Boston
Thomas Burrow f

George Cardona, Philadelphia
Neville E. Collinge, Manchester
Warren Cowgill }

Joseph De Chiccis, Philadelphia
George Dunkel, Zürich

Isidore Dyen, Yale University — Honolulu
Murray B. Emeneau, Berkeley
Bernhard Forssman, Erlangen
Robert À. Fowkes, New York
Paul Friedrich, Chicago

Eric P. Hamp, Chicago

Hans Henrich Hock, Urbana
Karl Hoffmann, Erlangen

Peter E. Hook, Ann Arbor
Stephanie W. Jamison, Cambridge Mass.
Jay H. Jasanoff, Ithaca N.Y.
Sara E. Kimball, Philadelphia
Jared S. Klein, Athens Ga.
Konrad Koerner, Ottawa
Frederik Kortlandt, Leiden
Winfred P. Lehmann, Austin
Giulio C. Lepschy, Reading
Albert L. Lloyd, Philadelphia
Olivier Masson, Paris

Francine Mawet, Bruxelles

M.A. Mehendale, Puna

Michael Meier-Brügger, Hamburg
Katrina Mickey, London

Anna Morpurgo Davies, Oxford
William G. Moulton, Princeton
Hugo Miihlestein, Neuchatel
Jean-Claude Muller, Bonn

Günter Neumann, Würzburg
Manfred Peters, Namur

Edgar C. Polomé, Austin

Jaan Puhvel, Los Angeles

Paolo Ramat, Pavia

Wilhelm Rau, Marburg

Ernst Risch, Zürich

Jochem Schindler, Wien

William R. Schmalstieg, Philadelphia
Hanns-Peter Schmidt, Los Angeles
Rüdiger Schmitt, Saarbrücken
Leslie Seiffert, Oxford

Andrew L. Sihler, Madison
Michael Silverstein, Chicago
Carlo de Simone, Tübingen

Otto Springer, Philadelphia

Klaus Strunk, München

K.M. Tiwary, Patna

Paul Thieme, Tübingen

Jürgen Untermann, Köln

Calvert Watkins, Cambridge Mass.
Horst Weinstock, Aachen

Werner Winter, Kiel

Gerd Wolandt, Aachen

Archicles Ladislaus Zgusta, Urbana
Arlene R.K. Zide, Chicago
Norman H. Zide, Chicago

VIII

Faculté des Lettres, Départment des Sciences de l'Antiquité, Bibliothéque, Université de
Genéve

Indogermanisches Seminar, Universitát Zürich

Indogermanistik, Teilbibliothek 6, Universität Erlangen-Nürnberg

Institut for lingvistik, Lingvistisk bibliotek, Kóbenhavns universitet

Institut für Sprachwissenschaft der Universitàt Innsbruck

Institut für Sprachwissenschaft der Universität Wien

Institut für Sprachwissenschaft der Universität zu Köln

Istituto di Linguistica e di Lingue Orientali, Università degli studi di Firenze

Lehrstuhl für Vergleichende Sprachwissenschaft der Universität Würzburg

Main Library, University of Birmingham

Neuphilologische Fakultät, Bibliothek, Universität Tübingen

Phonetisches Institut der Universität Hamburg

Seminar für Kultur und Geschichte Indiens, Universität Hamburg

Seminar für Vergleichende und Indogermanische Sprachwissenschaft, Freie Universität
Berlin i

Seminar für Vergleichende Sprachwissenschaft der Universität Würzburg

Sprachwissenschaftliches Seminar der Universität Marburg

Summer Institute of Linguistics, Ukarumpa Via Lae, Papua New Guinea

Universität Bayreuth, Universitätsbibliothek

Université Libre de Bruxelles, Bibliothèque

Universitetsbiblioteket i Oslo

University of Georgia Libraries, Athens Ga.

CONTENTS

POIGWOI. 1250 pu Ee rEG Ee RA ES ERR ERA GAME Moa ERRE V
UOI CFR M" DE VII
Bibliography of Henry M. Hoenigswald ............................... XII
Francisco R. Adrados (Universidad Complutense de Madrid)

Binary and multiple oppositions in the history of Indo-European . ........... 1
William Sidney Allen (Trinity College, Cambridge)

Syllabic prominence in ancient Greek: a typological approach .............. 11
Yoél L. Arbeitman (Bronx, New York)

Hittite pdi-, why no *wai-: an Anatolian-Indo-European heterogloss . ......... 19
Harold W. Bailey (Queens’ College, Cambridge)

Iranica in Caucasian o ioana ea i bb e teet eR eram und as x esi b e AE s 33

Philip Baldi (The Pennsylvania State University)
Prefixal negation of English adjectives: Psycholinguistic dimensions

of productivity cesoie rita Pr Re Ne DSTI Ne iosa 37
Robert S.P. Beekes (University of Leiden)

Indo-Buropean neutersin ......................,,............ 45
Thomas Burrow {+ (Balliol College, Oxford)

Four contributions to Sanskrit etymology ........................... 57
George Cardona (University of Pennsylvania)

On Sanskrit bhunäkti ‘aids, serves, protects’.......................... 65
Neville E. Collinge (University of Manchester)

Who did discover the law of the palatals? ........................... 73
Warren Cowgill t (Yale University)

The second plural of the Umbrian verb ............................. 81

George Dunkel (Princeton University)
heres,xnpwoTal: indogermanische Richtersprache ..................... 91

Isidore Dyen (Yale University and the University of Hawaïi)
Genetic classification in linguistics and biology . ....................... 101

Murray B. Emeneau (University of California, Berkeley)
Some notes on Dravidian intensives ............................... 109

Bernhard Forssman (Universität Erlangen-Nürnberg)
Vedisch ENEE EE EE E 115

Robert A. Fowkes (New York University)

Brythonic gender reduction — the Cornish picture ..................... 121
Paul Friedrich (University of Chicago)

The Proto-Indo-European adpreps (Spatio-temporal auxiliaries)............. 131
Hans Henrich Hock (University of Illinois)

Regular contact dissimilation een A NEE NEEN na ss 143
Peter E. Hook (University of Michigan)

Linguistic areas: getting at the grain of history ........................ 155

Stephanie W. Jamison (Cambridge, Massachusetts)
Mantra glosses in the Satapatha Bràhmana: more light on the development of the

Vedic verbal system pai ea aan re names 169
Jay H. Jasanoff (Cornell University)

The tenses of the Latin perfect system. «2s eoa ae ed ess 177
Sara Kimball (Philadelphia, Pennsylvania)

FH 3.4in EE 185

Jared S. Klein (University of Georgia)
The two senses of the term ‘anaphora’ and their functional unity:
evidence from the Rigveda 50e inerte pre ea na ava 193

Konrad Koerner (University of Ottawa)
The importance of Saussure’s M&moire’ in the development of

historical LINGUISTICS wc ene oS aw ee oe a we Roles eee we A 201
Frederik Kortlandt (University of Leiden)
Archaic ablaut patterns in the Vedic verb ........................... 219

Winfried P. Lehmann (University of Texas)
Theoretical views affecting successive reconstructions of the phonological system of

Protoslndo«EUrOpedie cac Spe eos RUP aa 229
Giulio C. Lepschy (University of Reading)

L'articolo indeterminativo (note per la storia della grammatica italiana) ....... 237
Albert L. Lloyd (University of Pennsylvania)

Old High German 4-, Old English @-,: a problem that won't go away ......... 243
Olivier Masson (Université de Paris X, Nanterre, et Ecole des Hautes Etudes, Paris)

Quelques noms grecs à l’Agora d'Athènes ........................... 255
M.A. Mehendale (Bhandarkar Oriental Research Institute, Pune)

Some E EE EE EE 261
Anna Morpurgo Davies (Somerville College, Oxford)

Folk-inguistics and the Greek word ............................... 263

William G. Moulton (Princeton University)
On Vowel Eeugthan Gallio us erm rre bcc EARN S RS QE ES UR 281

Hugo Mühlestein (Université de Neuchátel)
Ein Halbvers und einige Epitheta aus vorhomerischez Dichtung ............. 293

Edgar C. Polomé (University of Texas)

Initial PIE:*pWhk-in Germano c.ca 303
Jaan Puhvel (University of California, Los Angeles)

All our yesterdays eneo resco rana a aa eai E e olia ters Sad erige PE 315
Paolo Ramat (Istituto di Glottologia, Università di Pavia)

Verbi forti e verbi deboli in Germanico ............................. 319
Ernst Risch (Universität Zürich)

Sonderlall Griechisch? 7 #22. 22212 53.22 E RU A AIA 329.
Jochem Schindler (Universität Wien)

Zur avestischen Kompositionslehre: af.- ‘groB’ 2... ee 337
William R. Schmalstieg (The Pennsylvania State University)

The multiple origin of the Indo-European nominative case ................ 349
Hanns-Peter Schmidt (University of California, Los Angeles)

An Indo-Iranian etymological kaleidoscope .....................,..... 355
Rüdiger Schmitt (Universität des Saarlandes)

Altpersisch m-n-u-vi-i- = manauvië ................................ 363
Andrew L. Sihler (University of Wisconsin)

Further evidence in support of Brugman’slaw ........................ 367

Otto Springer (University of Pennsylvania)
Greek als, Latin *balan, Old High German bal (?) ‘marked by a blaze’:

a horse fanciers’ multilingual symposium ............................ 375
Klaus Strunk (Universität München) |

Further evidence for diachronic selection: Ved. rasti, Lat. regit etc. ......... 385
K.M. Tiwary (Patna University)

tulyasyaprayatnam savarnam Astàádhyayi (1.1.9) ...................... 393
Calvert Watkins (Harvard University)

Two Anatolian forms: Palaic askumauwa-, Cuneiform Luvian wa-a-ar-a ....... 399
Werner Winter (Universitàt Kiel)

Old Indic sunt-, Greek huids ‘son’ ................................ 405

Archicles Zgusta (University of Illinois)
Inscriptionis palaeo-Osseticae apud Zelenčuk flumen repertae lectiones quaedam
NOVAE:Proponnini: i suina rs Dr ans. gu 409

Arlene R.K. Zide (Loop College, C.C.C. and University of Chicago) and
Norman H. Zide (University of Chicago)

A KM Laryngeal as a conditioning factor for s-loss in Sora-Juray-Gorum ....... 417
Eric P. Hamp (University of Chicago)

Olr. ‘tab(ajir ‘brings”, ‘fait ‘comes’ ................................ 433

Bibliography of Henry M. Hoenigswald

A. Books.

1945

1960

1970

1973

1979

Spoken Hindustani, basic course. Volumes I, II (New York: Henry Holt), pp. viii,
169; ix—x, 171-433. (Identical with the edition prepared for the United States
Armed Forces Institute, Madison, Wisconsin, and published for the United States
Armed Forces by the Linguistic Society of America and the Intensive Language
Program of the American Council of Learned Societies.)

Language change and linguistic reconstruction (Chicago: University of Chicago
Press) pp. viii, 168, ?1965.

(ed. with CG Cardona and A. Senn) Indo-European and Indo-Europeans (= Haney
Foundation Series, 9.) (Philadelphia: University of Pennsylvania Press) pp. 440.
Studies in formal historical linguistics (Dordrecht: Reidel) (= Formal Linguistics
Series, edited by Henry. Hiz, volume 3.) pp. xiii, 63.

(ed.:) The European background of American linguistics (Lisse: Foris).

B. Articles and reviews.

1937

1938

1939
1940
1941

1942

‘’Ertopxos’, Studi Italiani di Filologia Classica 14: 83—87. i

'Su alcuni caratteri della derivazione e della composizione nominale indoeuropea',
Rendiconti, Istituto Lombardo, Lettere, n.s., 1: 267-274.

“Entypunos’, Studi Italiani di Filologia Classica 15: 87.

‘Problemi di linguistica umbra — a proposito delle Tabulae Iguvinae editae a Iacobo
Devoto’, Rivista di Filologia Classica 16: 274—294.

Rev.: i. Rosenzweig, Ritual and cults of Pre-Roman Iguvium, in: Rivista di Filologia
Classica 16: 427—429.

‘Studi sulla punteggiatura nei testi etruschi’, Studi Etruschi 12: 169—217.

‘Tlav-compounds in early Greek”, Language 16: 183—187.

‘On Etruscan and Latin month-names’, American Journal of Philology 62: 199—

206.

‘Campanian Inscriptions at Yale’, American Journal of Archaeology 48: 582—586.

‘The Epicharmian title Adyos Kai Aoyıva’, Language 17: 247—249.

Rev.: P.S.Costas, An outline of the history of the Greek language, with particular
emphasis on the Koine and the subsequent periods, in: Language 17:
262—263.

H.S. Deferrari, Outline of a theory of linguistic change, in: Classical World
34: 283.

"Three inscriptions in the University Museum, Philadelphia', American Journal of

Archaeology 46: 532-537 [with Edith Hall Dohan].

‘Philology’, Americana Annual.

XIV

1943

1944

1945

1946

1947

1948

1949

1950

1951

Rev.: E.H. Warmington, Remains of Old Latin, IV, in: Language 18: 65-66.
J.R. Ballantyne, First lessons in Sanskrit grammar, revised by L.A. Ware, in:
Classical World 35: 248.

‘Philology’, Americana Annual.

Rev.: V. Georgiev, Vorgriechische Sprachwissenschaft, I, in: Language 19: 269—
272,

‘Internal reconstruction”, Studies in Linguistics 2 (1943-4): 78-87.

‘Philology’, Americana Annual,

Rev.: K.L. Pike, Phonetics, in: Journal of the American Oriental Society 64:151—
185;
‘Philology’, Americana Annual.

Rev.: R.A. Hall, Hungarian Grammar, in: Studies in Linguistics 3: 46—50.

‘Sound change and linguistic structure’, Language 22: 138-143. [Reprinted in
Readings in linguistics I, edited by M. Joos (New York: American Council of Learn-
ed Societies, 1957, Chicago: University of Chicago Press, 1958), pp. 139—141.]
‘The phonology of Etrusco-Roman names”, Transactions of the American Philologi-
cal Association 77: 319.

‘Etruscan’, in: Encyclopedia of literature, edited by J.T. Shipley (New York:
Philosophical Library), 1; 278—279.

‘Philology’, Americana Annual.

Rev.: M.B. Emeneau, Kota texts I, in: Journal of the American Oriental Society
66: 187—188.
*On Varro, De Lingua Latina, V. 15°, American Journal of Philology 68: 198—199.

Rev.: E.H. Sturtevant, An introduction to linguistic science, in: Language 23:
437—442.

*Declension and nasalization in Hindustani', Journal of the American Oriental

Society 68: 139—144.

*Diachronic sound charts', Studies in Linguistics 6: 81—94.

Rev.: V. Georgiev, Vorgriechische Sprachwissenschaft, Il, in: Language 24: 198—
199.
H. Fournier, Les verbes “dire” en grec ancien (exemple de conjugaison
supplétive),in: Language 24: 199—205.
L.R. Palmer, A grammar of the post-Ptolemaic papyri: volume 1, Accidence
and word-formation; part 1, the suffixes, in: Language 24: 205--212.
*Antevocalic u-diphthongs in Latin’, Language 25: 392—394.
‘A note on Latin prosody: initial s impure after short vowel', Transactions of the
American Philological Association 80: 271—278.
‘The principal step in comparative grammar’, Language 26: 357--364. [Reprinted in
Readings in linguistics I, pp. 298-302.]
‘Morpheme order diagrams’, Studies in Linguistics 8: 79—81.

Rev.: R.E. Chandler and A.R. Hefler, Handbook of comparative grammar for
students of foreign languages, in: Studies in Linguistics 8: 40.

Rev.: H. Krahe, Die Indogermanisierung Griechenlands und Italiens, in: Language
27:374-375.

1952

1953

1954

1955
1956

1957

1958

1959
1960

1961

1962

1963

South Etruscan and Cypriote writing’, American Journal or Archaeology 56;174.
‘Laryngeals and s movable', Language 28: 182—185.
‘The phonology of dialect borrowing’, Studies in Linguistics 10: 1—5.

Rev.: M.B. Emeneau, Studies in Vietnamese ( Annamese) grammar, in: Studies in
Linguistics 10: 96—98.

“PA, AEAAE, AAZT®, and the semivowels', Language 29: 288—292.

‘IT Fondamenti della storia linguistica e le posizioni neogrammatiche', Lingua Nostra

12: 47-50.

Rev.: R.H. Robins, Ancient and mediaeval grammatical theory in Europe, in: Lan-
guage 29:180—182.
F. Slotty, Beitráge zur Etruskologie, in: Language 29:183—186.
"Media, Neutrum und Zirkumflex', in: Sprachgeschichte und Wortbedeutung, Fest-
schrift A. Debrunner (Bern: Francke), pp. 209—212.
‘Linguistics in the sixteenth century’, Library Chronicle 20: 1-4.

Rev.: W.P. Lehmann, Proto-Indo-European phonology, in: Language 30:468—474.

‘Change, analogic and semantic’, Indian Linguistics 16: 233—236.

Rev.: P. Thieme, Die Heimat der indogermanischen Gemeinsprache, in: Language
32:313--316.

Rev.: A. Martinet, Economie des changements phonétiques, in: Language 33:
575—583. m e

‘A Latin trace of the construction data radhamsi’, Indian Linguistics 19—20 (1958—

9 [Sir Ralph L. Turner Jubilee volume]): 232—234.

. Rev.: Taraporewala memorial volume, in: Language 34:516—518.

J. Whatmough, Poetic, scientific, and other forms of discourse. A new
approach to Greek and Latin literature, in: American Journal of Philology
79: 290-293.

‘Some uses of nothing’, Language 35: 409—420.

‘Phonetic similarity and internal reconstruction', Language 36: 191—192.

Rev.: F. de Saussure, Course in general linguistics [translated by W. Baskin], in:
Classical World 53: 267.

Rev.: J.W. Poultney, The bronze tablets of Iguvium, in: American Journal of
Philology 82: 301—306.

‘Bilingualism, presumable bilingualism, and diachrony’, Anthropological Linguistics

4: 1—5.

‘Synchronic and diachronic aspects of distributional meaning’, in: Georgetown

Monographs in Language and Linguistics 13: 21-28.

‘Lexicography and grammar’, in: Problems in lexicography, edited by F.W. House-

holder and S. Saporta (Bloomington: Indiana University Press), pp. 103—110.

‘Notes on change and reconstruction’, International Journal of American Linguistics

28: 80-84.

Rev.: A.C. Moorhouse, Studies in the Greek negatives, in: Language 38:310—313.
C. Schick, J! linguaggio. Natura, struttura, storicità del fatto linguistico, in:
Romance Philology 15: 333—335.

‘On the history of the comparative method’, Anthropological Linguistics 5: 1—11.

‘A contribution to Indo-Iranian phonology: the voiceless aspirates’, in: Transactions

of the Twenty-fifth International Congress of Orientalists 2: 323--326.

XVI

1964

1965

1966

1967

1968

1969

1971

‘Are there universals of linguistic change?”, in: Universals of language, edited by J.
Greenberg (2nd edition, Cambridge, Mass.: MIT Press), pp. 30—52.

Rey.: J. Gagnepain, Les noms grecs en -os et en -2, in: Language 39: 107—110.
‘People and language’, Journal of General Education 16: 203—214.

‘Mycenaean augments and the language of poetry’, in: Mycenaean studies. Proceed-
ings of the Third International Colloquium of Mycenaean Studies, held at “Wing-
spread”, 4-8 September 1961, edited by E.L. Bennett, Jr. (Madison: University of
Wisconsin Press), pp. 179—182.

‘Graduality, sporadicity, and the minor sound change processes’, Phonetica 11:
202-215.

‘Allophones, allomorphs, and conditioned change’, in: Proceedings of the Ninth
International Congress of Linguistics, edited by H. Lunt (The Hague: Mouton),
pp. 645—649.

‘George Melville Bolling', Language 40: 329 —336.

Rev.: J.T. Waterman, Perspectives in linguistics, in: American Anthropologist 66:
478—479.

‘Indo-Iranian evidence’, Evidence for laryngeals, edited by W. Winter (The Hague:

Mouton), pp. 93—99.

‘A property of “Grassmann’s Law” in Indic’, Journal of the American Oriental

Society 85: 59--60.

‘Phonetic reconstruction’, in: Proceedings of the Fifth International Congress of

Phonetic Sciences, edited by E. Zwirner (Basel: Karger), pp. 25—42.

Rev.: J. Lyons, Structural semantics: An analysis of part of the vocabulary of
Plato, in: Journal of Linguistics 1: 191—196.

*Criteria for the subgrouping of languages', Ancient Indo-European dialects, edited

by H. Birnbaum and J. Puhvel (Berkeley, Los Angeles: University of California

Press), pp. 1—12.

‘A proposal for the study of folk-linguistics', Sociolinguistics, Proceedings of the

UCLA Sociolinguistics Conference, 1964, edited by W. Bright (The Hague: Mou-

ton), pp. 16—26.

‘Champion or star?', in: Papers in linguistics in honor of Léon Dostert, edited by

W.M. Austin (The Hague: Mouton), pp. 106—107.

‘The syllabaries and Etruscan writing”, Incunabula Graeca 25: 410—416.

*Certain semivowel sequences in Greek', in: Pratidanam; Indic, Iranian, and Indo-

European studies presented to Franciscus Bernardus Jacobus Kuiper on his sixtieth

birthday, edited by J.C. Heesterman, G.H. Schokker, C.I. Subramonian (The

Hague: Mouton), pp. 20--23.

‘A note on overlength in Greek', Word 24: 252—254.

Rev.: T.A. Sebeok, editor, Portraits of linguists: A biographical source book for
the history of western linguistics, in: American Anthropologist 70: 422.
G. Stern, Meaning and change of meaning in English, in: Foundations of
Language 4: 101-104.
‘Our own family of languages’, in: Linguistics today, edited by A.A. Hill (New
York: Basic Books), pp. 59-70.
‘Language history and creole studies’, in: Pidginization and creolization of lan-
guages, edited by D.H. Hymes (Cambridge: Cambridge University Press), pp. 473—
480.

1972

1973

1974

1975

XVII

Rev.: N. Chomsky, Cartesian linguistics, M. Leroy, Main trends in modern lin-
guistics; R.H. Robins, 4 short history of linguistics, in: Journal of the
History of the Behavioral Sciences 7: 206—209.

W.P. Lehmann, A reader in nineteenth century historical Indo-European
linguistics, in: Lingua 26: 423—427.
A.J. Pfiffig, Ein Opfergelübde an die etruskische Minerva, in: American
Journal of Archaeology 71: 2277 —228.
E. and K. Zwirner, Grundfragen der Phonometrie, in: Language 47: 189—
190.
‘A doubtful note on Greek relative chronology’, in: Studies in linguistics in honor
of George L. Trager, edited by E. Smith (The Hague: Mouton), pp. 231—232.

Rev.: M. Cristofani, editor, Corpus Inscriptionum Etruscarum, 2.1.4, in: American

Journal of Archaeology 76: 452—453.

O. Szemerényi, Einführung in die vergleichende Sprachwissenschaft, in:

American Classical Review 2: 153—154.

W.F. Wyatt, Jr., Metrical lengthening in Homer, in: Language 48: 929—940.
‘Introductory remarks’, in: Current trends in linguistics, volume 11, edited by T. A.
Sebeok, H.M. Hoenigswald, R.E. Longacre (The Hague: Mouton), pp. 45-49.

‘The comparative method’, in: Current trends in linguistics, volume 11, pp. 51- -62.
‘Linguistics’, in: Dictionary of the history of ideas, edited by P. Wiener, volume 3,
pp. 61-73.

‘Indo-European *p in Celtic and the claim for relative chronologies’, Journal of
Indo-European Studies 1:324—329.

‘Relative chronology --- Notes on so-called intermediate stages’, in: Proceedings of
the XIth International Congress of Linguists, edited by L. Heilmann (Bologna: Il
Molino), volume 1, pp. 369—373.

‘Stefán Einarsson (1897—1972)’, in: Year Book of the American Philosophical
Society 1973: 120—122.

Rev.: W.M. Cowan, Workbook in comparative reconstruction, in: General Lin-
guistics 13: 99—101.

‘Internal reconstruction and context’, in: Historical linguistics: Proceedings of the

First International Conference on Historical Linguistics, edited by J.M. Anderson,

C. Jones (Amsterdam: North Holland Publishing Company, New York: American

Elsevier), volume II, pp. 189—201.

‘Fallacies in the history of linguistics: Notes on the appraisal of the nineteenth

century’, Studies in the history of linguistics: Traditions and paradigms, edited by

D.H. Hymes (Bloomington: Indiana University Press), pp. 346. .358.

"Alfred Ernout (1879—1973Y, Bulletin, Linguistic Society of America, 61: 38.

Rev.: Heller, Communicational analysis, in: Historical Review 1974: 1137.

*Su -nd- latino', Archivio Glottologico Italiano 60: 55—58.

‘On Schleicher’s tree and its trunk’, in: Ut Videam: Contributions to an under-
standing of linguistics: for Pieter Verburg on the occasion of his 70th Birthday,
edited by W. Abraham (Lisse: Peter de Ridder Press), pp. 91--93.

‘Sarah C. Gudschinsky (1919-1975)’, Bulletin, Linguistic Society of America,
66.16.

XVII

1976

1977

1978

Rev.: W.A. Benware, The study of Indo-European vocalism in the 19th century,
from the beginnings to Whitney and Scherer, in: Historiographia Linguistica
2: 238—241.
J.A. Kerns and B. Schwartz, A sketch of the Indo-European finite verb, in:
Journal of the American Oriental Society 95: 147—148.
P. Schachter and Fe. T. Otanes, Tagalog reference grammar, in: Journal of
the American Oriental Society 95: 148.
E. Seebold, Das System der indogermanischen Halbvokale, in: Journal of
the American Oriental Society 95: 543—544.
M.L. Samuels, Linguistic evolution with special reference to English, in:
Language 51: 457—461.
T.A. Sebeok, editor. Current trends in linguistics, volume 10, in: Language
Sciences 37: 35—36.
‘Summary commentary', Oral literature and the formula, edited by B.S. Siolz and
R.S. Shannon (Ann Arbor: Center for the Coordination of Ancient and Modern
Studies, University of Michigan), pp. 273—278.

Rev.: W.A. Benware, The study of Indo-European vocalism in the 19th century,
from the beginnings to Whitney and Scherer, in: Journal of the American
Oriental Society 96: 137.

H. Berger, Das Yasin-Burushaski, in: Journal of the American Oriental
Society 96: 137.

A. Pott, Einleitung in die allgemeine Sprachwissenschaft, in: Journal of the
American Oriental Society 96: 137.

B. Delbrück, Introduction to the study of language [new edition of the
English translation by E. Channing], in: Journal of the American Oriental
Society 96: 137--138.

‘Intentions, assumptions, and contradictions in historical linguistics’, in: Current

issues in linguistics theory, edited by R.W. Cole (Bloomington: Indiana University

Press), pp. 168—193.

‘Diminutives and tatpurusas: the Indo-European trend toward endocentricity',

Journal of Indo-European Studies 5: 9--13.

Rev.: G. Nagy, Comparative studies in Greek and Indic meter, in: American Jour-
nal of Philology 98: 82—88.

‘An aspect of semivowel asymmetry in Indo-European’, Bulletin of the Department

of Comparative Philology and Linguistics, University of Calcutta 3, (S.K. Chatterji

Memorial Volume), pp. 14—18.

‘A note on Rgvedic bahuvrihis’, Indian Linguistics 39 (P. B. Pandit Memorial Vo-

lume): 29—30.

‘Adjectives as first compound members in Homer’, in: Linguistic and literary

studies in honor of Archibald A. Hill, edited by M.A. Jazayery, E.C. Polomé, and

W. Winter (The Hague: Mouton), volume III: Historical and comparative linguistics,

pp.91-95.

‘Secondary split, typology, and universals’, in: Recent developments in historical

. phonology, edited by J. Fisiak (The Hague: Mouton), pp. 173—182.

‘The annus mirabilis 1876 and posterity', Transactions of the Philological Society,
London, 1978: 17—35.

1980

1981

1983

1984

1985

XIX

‘The impact of vowel syncope in Latin’, in: /talic and Romance: Linguistic studies
in honor of Ernst Pulgram, edited by H.J. Izzo (Amsterdam: Benjamins), pp. 53—
58.

‘Notes on reconstruction, word order, and stress’, in: Linguistic reconstruction and
Indo-European syntax, edited by P. Ramat (Amsterdam: Benjamins), pp. 69—87.
“Initial semivowel clusters’, in: Studien zur Indologie und Iranistik 5/6 (= Festschrift
Paul Thieme),edited by G. Buddruss, A. Wezler, pp. 83--86.

‘The name of the Argolid’, Word 31: 105—107.

‘A reconstruction’, in: First person singular, edited by B.H. Davis and R. O’Cain
(Amsterdam: Benjamins), pp. 22—28.

‘Degrees of genetic relatedness among languages’, in: Suniti Kumar Chatterji com-
memoration volume, edited by S.P. Mallik (Burdwan: University of Burdwan), pp.
113-115.

Rev.: Language and linguistic area. Essays by Murray Emeneau. Selected and in-
troduced by Anwar S. Dil, in: Journal of the American Oriental Society
101: 238—239.
B. Brogyanyi, editor, Studies in diachronic, synchronic, and typological lin-
guistics. Festschrift for Oswald Szemerényi on the Occasion of his 65th
Birthday, in: Kratylos 25 (1980 [1981]): 95—104.
‘Doublets’, in: Essays in honor of Charles F. Hockett, edited by F.B. Agard, G. Kel-
ley, A. Makkai, V.B. Makkai (Leiden: Brill), pp. 167—171.
*Etymology against grammar in the early 19th century’, in: Histoire, Epistémologie,
Langage 6.2, pp. 94—100.
*Apropos of Greek word-accent', in: Studi linguistici e filologici per Carlo Alberto
Mastrelli (Pisa: Pacini), pp. 219—224.
"Distinzioni reali et distinzioni chimeriche nella classificazione dei cambiamenti
fonologici', in: Società Linguistica Italiana: XVI? Congresso Internazionale di
Studi: Linguistica storica e cambiamento linguistico, pp. 111—118.
‘Sir William Jones and historiography’, in: For Gordon Fairbanks, edited by V.Z.
Abson and R.L. Leed (Oceanic Linguistics Special Publication No. 20; University of
Hawaii Press), pp. 64—66.
"From plain to worse', in: Sprachwissenschaftliche Forschungen, Festschrift für Jo-
hann Knobloch, edited by H.M. Olberg, Gernot Schmidt, and H. Bothien (Inns-
brucker Beitráge zur Kulturwissenschaft, Band 23; Innsbruck: AMOE), pp. 167—
170.

Rev.: P. Baldi, An introduction to the Indo-European languages, in: General Lin-
guistics 24:71—72.

G. and L. Bonfante, The Etruscan language. Àn introduction, in: Classical
Review 24:340—341.

Binary and multiple oppositions in the history of Indo-European

Francisco R. Adrados (Universidad Complutense de Madrid)

1. This paper is written on the strength of a series of ideas which I have attempted to
justify elsewhere and which I am here obliged to take simply as a starting-point.’

In fact it deals with the idea that the development of Indo-European took place in three
stages, which may be defined schematically as follows:

Stage 1: Pre-flexional Indo-European or Proto-Indo-European (IE 1)

This functioned on the basis of root-words, either nominal-verbal or pronominal-adverbial
ones, which determine each other to make up syntagms and sentences through word-order,
accent placing and certain enlargements, albeit without proper inflexion and without the
later categories of Indo-European having yet arisen.

Stage 2: Monothematic Indo-Eurpean (IE 2)

This phase, preserved in Anatolian (which was left isolated to the south of the Caucasus)
and in archaisms in other languages, already has inflexion although always on one stem. In
the noun, it opposes singular and plural, animate and inanimate, and has a case system; in
the pronoun, it opposes persons; in the verb, singular and plural, three persons, indicative
and imperative, present and preterite, active and middle voice. These oppositions are
basically expressed by means of oppositions between series of endings. There is the begin-
ning of stem opposition: basic and derived verbs (causative, iterative, etc.); sg. and pl. Nom.-
Acc.-Voc. n. (-om/-a, -à); heteroclitic nouns.

Stage 3: Polythematic Indo-European (IE 3)

The former oppositions of categories are maintained in this phase, but some of them are
extended: that of gender is now between masculine, feminine and neuter, that of number
between singular, plural and dual (but not always), that of tense between present, preterite
and (at times) future, that of mood between indicative, imperative, optative and (in certain
languages) subjunctive, that of voice between active, middle and (at times, too) passive.
These extensions now need stem oppositions. And the same goes for other new oppositions:
that of the degrees of comparison in the adjective and that of the aspects in the verb. An
old opposition, that of present and preterite, now not only opposes endings but also stems.
This type of Indo-European is the one which is the object of traditional description.

According to this conception, there has been progressive complication of the oppositive
systems; on the other hand, there have been implications between them, that is, one form

may intervene in more than one system of oppositions (gender, number and case; tense,
aspect, voice, number and person). It is, moreover, curious that all the oppositions in Indo-
European 2 are binary, with the exception of those of person and case; all are basically
carried out with the help of endings, although there are also oppositions between lexical
items (nouns which are only neuter or only animate, for example), and, to a minimal
extent, between stems (in the above-mentioned cases). In Indo-European 3, on the other
hand, although it still preserves the old formal resources of the oppositions, the new
resource of stem opposition prevails, either by serving the old oppositions or by extending
them or by creating other new ones.

An opposition, in fact, may make use of diverse formal resources, according to its antiquity
or the category or function it refers to, and may use one or another or a combination of
both.

The remarkable thing, however, is that at the time when the grammatical resources used
(the lexical ones existed alongside these) were endings, the oppositions were binary, with
one or two exceptions; and when these resources were complemented by other grammatical
ones, based on the opposition of stems, they became multiple, being generally ternary. The
growth of Indo-European morphology was therefore thus tied up with the above mentioned
formal resources and obviously had certain repercussions at the level of content. Yet these
repercussions may be summed up as the passing from binary oppositions to ternary ones or
even to those with more terms.

Therefore it is a matter of explaining this evolution and the reasons for it. It also means
explaining the great exception which existed: the case system which in Hittite has seven
terms and in other languages a variable number of them, e.g. a maximum of eight, as is
known, in Indo-Iranian.

I believe that a further argument should be added to those I have already given in the
above-mentioned publications, in favour of the hypothesis that the binary oppositions are
the more archaic and that the grammatical resource based on the opposition of endings is
also the most archaic. This argument would be of a general type: grammatical oppositions
generally begin as binary through the opposition of two terms (two words, two endings or
various grammatical elements). It is clear that the opposition of two words is not yet a
grammatical fact; neither are the other oppositions when they are isolated. A series of
opposed terms are needed: *patér/*matér, *bhräter /*suesör ...; -ti/-t, -si/-s; *line kFti/
“leloikke, *krneuti / *klektore . . .

On the other hand, the last examples indicate that in the oppositions expressed by gram-
matical elements one must count upon allomorphism, as one must likewise count upon
defectiveness (*uoide, without a present), syncretism, amalgams, neutralizations and other
irregularities. As I have explained elsewhere?, these irregularities shed light upon the in-
variably secondary nature of the oppositions of inflexional Indo-European (2 and 3).

If this is so, it would appear that the multiple oppositions must forcibly have originated in
an even more secondary way: through the split of one term of a binary opposition (the
animate split into masculine and feminine); through the addition of a new term which
opposes one of the basic binary opposition; or through combinations within a complex
system of independent binary systems.

Before going further into this subject, however, I have a few things to say on the origin of
oppositions in general: that is, in fact, on the origin of the binary systems of oppositions

which are the basic ones. I believe that this has been acknowledged whenever a diachronic
criterion has been applied which was at the same time structural, in order to explain how
oppositions appear in one stage of a language which did not exist in a former one; and how
formal elements which could not express oppositions which did not yet exist in the former
stage could then express them in the later stage. I refer briefly to papers by Kurytowicz?
and others of mine^ in which we refer to precedents of these ideas in A. Martinet and
others.

Very briefly:

a) A lexical, or lexico-grammatical item may be opposed as a negative term to the same
item when a second element is added to it, thus becoming the positive term. Thus,
a verbal stem followed by elements such as -m, -s, -t, -nt does not indicate tense; but if this
complex T-m (T-s, T-t, T-nt) is opposed to another T-m-i (T-s-i, T-t-i, T-nt-i) in which -i
signals ‘present’, it comes to mean ‘preterite’ through opposition.

b) Certain elements which originally had no meaning of their own acquired it in certain
contexts. For example, and -é are originally radical elements (in disyllabic roots}; but they
are at times defined as characteristics of the indicative (as against other of the subjunctive),
at others, they indicate the opposite; or else: 4 signals the indicative and -e the subjunctive
(Lat. amas/ames). As is today acknowledged, the same goes for the case of the thematic
vowel which marks indicative as against the long vowel of the subjunctive or subjunctive
as against athematic forms of the indicative, etc. As soon as one of these elements, ac-
companying a verb in an indicative context, is interpreted as expressing the indicative, it is
then used to this end in other different verbs; and the same occurs with the subjunctive.
Thus, Brugmann has already explained the conversion into a feminine marker of an 4
which was originally radical (in *¢/nd).

Naturally, the signifiers are later extended and complicated, but the basic pattern is the
following: there is always an underlying opposition between two terms, or two series of
words.

2. In the second part of this paper, I shall now examine the inventory of binary opposi-
tions in Indo-European, together with their means of expression: I shall give a few indica-
tions as to how they were used for the creation of these oppositions. Then I shall examine
the exceptions existing within Indo-European 2: ternary systems or those with more terms.
This will clear the way for the last part of the paper: the systems of multiple oppositions
(through extension of the binary ones or new creations) in Indo-European 3.

The binary oppositions of Indo-European 2 preserved in Anatolian are as follows: in the
noun: animate/inanimate; in the verb: indicative/imperative, present/preterite and active/
middle voice; in one and the other: singular/plural.

Of these oppositions, that of gender (animate/inanimate) and that of voice (active/middle)
have a lexical basis although this is lost in the gender of adjectives, which have “motion”.
To begin with the former, nouns are by definition ‘tantum’ forms (either only animate or
only inanimate); the concept of defectiveness is habitual, although it may at times be
lacking (cf. doubtlessly archaic oppositions of the type tree names in -os/fruit names in -om,
the types represented by Gr. yn/reöov, Lat. terra/tellus, O.I. agnis/Hitt. pahhur, Lat. aqua]
Gr. 06 cop). As is well-known, to the lexical item another grammatical one is added: animates
with different Nom., Acc. and Voc./inanimates with syncretic Nom.-Acc.-Voc.? . The theo-

ry I have expounded in other papers on the origin of the opposition is the following: a sub-
class of nouns incapable of performing the function of subject and the impressive-expres-
sive one were opposed to another sub-class of nouns capable of both functions: both types
also performed that of object (cf. Adrados 1975: 398 ff). The nouns of the first and second
class were organized into two series, endowed with suitable grammatical markers (although
the inanimates could also later function as Nominative and Vocative, they extended the old
Accusative form to these uses, whence the syncretism).

Thus, a binary lexical opposition was extended to two lexical series and was grammaticalized
with one positive and one negative term. Something similar may be said of the voice oppo-
sition: there are verbs which are only active, others with different roots which are only
middle voiced: they are the lexical basis of an opposition which was likewise gramma-
ticalized. In this case there was systematic opposition of two series of endings (in gender,
this occurred in Nominative and Vocative but not in Accusative). Verbal forms with -e/o
stem or with -e/o ending which could appear together with verbal forms of the middle
voice in themselves, spread as carriers of this voice; on the contrary, radical verbs without
-e/o or with endings not followed by -e/o, remained characterized as of the active voice.
Traces of the secondary nature of the opposition are, among others: the fact of their being
at times connected to lexical items still; the fact that stems with -e/o or endings with -e/o
may at times be of the active voice or indifferent to voice (Gr. pépet, Hitt. -ta and Gr. -opa
in the active voice: the perfect in -a being indifferent); the fact that there are ‘tantum’
forms which are only either active or middle voice; that the opposition did not spread to
languages such as Baltic and Slavic (although there is a theory, which I believe is erroneous,
that it was lost) (Autor 1981: 41 ff.).

Thus, binary lexical oppositions were the nucleus for binary oppositions of series of roots
provided with grammatical elements (endings or ® endings) either concommitant with
mentioned oppositions or used secondarily to express those. There comes a moment when
only the opposition of endings added to one and the same lexical item is sufficient.

There is a certain difference in the opposition of number (singular/plural). Here, too, there
are ‘tantum’ lexical forms (either only singular or only plural); but they are not opposed to
each other and there are, on the other hand, syncretic forms which are not yet differen-
tiated: this is what often occurs in Hittite inflexion. In number the opposition of endings,
when it is there, comes from secondary differentiations: this is how I have interpreted the
oppositions -os/-os (with metrical lengthening), -os/oi (with agglutination of -i), -s/-es (with
difference of vocalic degree), etc.; and -m/-ms, bhi/-bhis (with secondary interpretation of
-s as a plural marker). Probably, originally anumerical nouns were characterized as singular
or plural when they were determined by lexical items of these types (‘one’/‘two’, ‘three
..., T/‘we’...) (cf. Adrados 1975: 427 ff; 1984 b)

In the last instance, these are re-makes of lexical oppositions with the aid of endings. But in
others cases, an ending has also been used with its own value as against another @ ending
which has become the negative term: thus in the temporal opposition (-mi/-m, -si/-s . . A
it is necessary to point out that the negative term may act as a positive one (the type of Gr.
Tins, Hitt. arta as present forms) and that the - may be extended outside its limits. The
use of an ending to mark one term of an opposition as against the lack of ending to mark
the other is also found in the indicative/imperative opposition as it appears in Hittite
(although the negative term acquires certain formal markers secondarily).

To sum up: the basis of the binary oppositions is in lexical or grammatical pairs (specific
ending as against @) or in the abstraction that opposes "one" to the other numerals. Or in
the combination of different resources: at times, by attributing to both opposed terms
forms which were in principle indifferent to opposition.

3. What happens with the multiple oppositions of Anatolian can now be understood.

The first is that of person: first, second and third. Obviously, this is based on the existence
of the pronominal forms T’, ‘yov’ and the other pronouns and nouns in general. On the
other hand, it has no lexical basis: all verbs have the three forms.

However, the partial identity of the second and third persons singular is only too well-
known. In fact, in Greek, within the forms which were originally without endings, vépetc/
pépet were only secondarily differentiated; in Lithuanian something similar occurred, but
here the differentiation was carried out with the aid of vowel timbre (2nd -e < *-ei/3rd -a
< *-o). When there are endings, things are even clearer: in Hittite in the preterite we find
+, 3, -ta and -šta in both persons and there are parallels in other languages (cf. Adrados
1974: 621 ff., 635 ff.). In fact, second-third forms were secondarily split, certain ones
being used for the second, and others for the third; they were old allomorphs.

That is to say that the ternary opposition was carried out by means of a scission within one
of the two terms, the negative. Once more, we come to the same result. And if we consider
that the opposition existing between base verb/iterative/causative was a ternary opposition,
then the evolution which led to this result is here different: base verb/iterative and base
verb/causative are the two underlying binary oppositions which were then combined. In
each one of them a form considered to be non-characterized was opposed to another
characterized one following a pattern that is by now well-known to us. In this case it is
clear that an opposition between stems was used for the first time, which has been con-
sidered as a model for all later ones (cf. Rosenkranz 1979: 227).

I believe that this combination in a multiple opposition of older binary oppositions is what
happened as far as the case system is concerned. Naturally, I am not starting with the
theory of an older system of eight cases which were later syncretized here and there; in this
sense I would refer to my above-mentioned works and to Villar (1974), as likewise to Leh-
mann (1958: 179—202), Kurytowicz (1964: 190 ff), Fairbanks (1977: 101—131), Schmal-
stieg (1980: 73 ff). Today, there is a widespread opinion that the system of Nominative,
Accusative, Genitive and Vocative is older, as likewise a form variously interpreted either
as a Locative or as an extra-syntactic one; for Villar and myself, the Dative/Locative
split is secondary.

However, without going further into the creation of the system, as it basically already
existed in Indo-European 2,one should note that if we believe that it is the synthesis or
combination of various systems of binary oppositions, this is because each of them belongs
either to a different function of the language or to different contexts. I believe that there
were initially several oppositions:

a) A functional opposition: base noun/Vocative

It is clear that the vocative inherits a special use of the noun in expressive and impressive
non-representative function: it was formalized by means of junctures, initial accent placing,
the use of the pure stem, etc.

b) An Adverbal opposition (created within the base noun)

The Nominative/Accusative opposition refers to that existing between two actants or deter-
miners of the verb: this is its context. Of course in the Nominative there were still non-
casual uses (asyntactic Nominative pendens, etc.) and the Accusative at times acquired new
contexts secondarily (Accusative of relationship to the noun, adverbial Accusative). The
opposition is expressed by means of endings (including the @ ending), which in general
mark these two types of verb determination (Adrados 1975: 407 ff).

c) An adnominal opposition
The base noun/Genitive opposition opposes any noun to another one which has the function
of noun determiner. The Genitive as verb determiner is secondary as is generally believed.

d) A sentence opposition

The remains of the base noun which were not split into the three former binary systems
were used with a local value opposed to any other noun in the sentence. I shall not enter
here into a discussion of the traditional theory that there were originally four cases (Da-
tive, Locative, Instrumentalis and Ablative); I believe that the Dative/Locative scission is
secondary (their markers -(, -i and -ei are the same) and that the Ablative and Instrumentalis
were developed as dialectal. I refer to the above-mentioned publications to this end and for
all details concerning the formalization of the different oppositions.

It is clear that the evolution of Indoeuropean tended to confuse functions and distributions
and to combine the four binary oppositions in one single system. Opposition c), for exam-
ple, as from a certain moment onwards, opposed to the Genitive both the Nominative, the
Accusative, and even the Vocative and Locative; all cases come into the negative term
opposed to the Vocative, and, of course, into that opposed to the Locative.

Although very briefly expounded, I believe that only thus may the origin of the multi-
lateral case system be explained. Despite all, it has preserved traces of the following ar-
chaisms: the limitation of certain cases to certain distributions (or preference for them),
the opposition of certain cases to others (and not to all of them) also in certain distribu-
tions, systems of partial oppositions within the overall case system.

Thus, Indo-European 2 not only provides proof that the oldest systems are the binary ones,
and specifically those created with either lexical or ending elements, but that it was the
first which resources that the whole of Indo-European was thenceforth to handle in order
to create systems of multiple oppositions: either by splitting one term of an opposition or
by synthesizing several into one, and by using the opposition of stems as a morphological
resource as well as the former ones.

4. This brief account of the creation of systems of grammatical oppositions in Indo-Euro-
pean 2 may serve as a guide for tracing another parallel with regard to Indo-European 3: I
have already stated that there is an extension of the morphological resources (generalized
use of stem oppositions) and that at times this is an extension of a binary system in Indo-
European 2 which as a consequence becomes ternary or multiple, at others creates a new
system — here, too, as always, on the basis of a binary one.

There are a series of concepts which I shall handle — concepts which partly correspond to
already known data in Indo-European 2 — and which should be more precisely defined. I
speak of scission when one term of a binary opposition is split into two terms that are
opposed to each other: the allomorphic stems of which it consisted are shared by the two

terms of the secondary opposition now created. Binary opposition thus becomes ternary.
This is what already occurred in Indo-European 2 when the person opposed to the first was
split into second and third as we have seen.

The concept of allomorphization means that different stems become part of one and the
same term of an opposition that is created. Thus, this occurred already in Indo-European 2
in the case of the animates and the inanimates, for example.

I speak of combination when two or more binary oppositions with a common term are
combined into a multiple one: just as in the case system which already existed in Indo-
Furopean 2.

Finally, the concept of addition means that, at times, one or more terms of an opposition
are secondarily opposed to a new term, whatever its origin may be. The opposition thus
becomes ternary or multiple. This is somewhat parallel to what occurs when a term is
opposed to another to make up a binary opposition; it is therefore a continuation of the
same procedure.

Let us first briefly examine the extensions of the old binary oppositions of Indo-European
2 mentioned above. For more detailed analysis, I refer to the above-mentioned biblio-
graphy and to the already quoted works. We find:

a) Addition

This is produced with ending elements when the plural in a series of Indoeuropean lan-
guages becomes a negative term opposed to a newly created positive one, the dual, which
refers to certain numerical uses (two individuals or objects in certain circumstances), which
are also expressable as a negative term by the plural. With regard to the origin I suggest
analogy with *duo and *ambho in the case of -0/u) for the dual endings (cf. Adrados 1975:
440 ff).

b) Scission

Animate nouns split into masculine and feminine: certain lexical items and often stems
followed by certain grammatical elements (above all masc. -o/fem. -2) form two opposed
series by means of a process of allomorphization. As is known, the feminine is a positive
term and the masculine a negative one. On the marking of masculine and feminine (respec-
tively) with elements such as -o and -z, which at first were indifferent to this distinction,
several hypotheses may be formulated (cf. Adrados 1975: 481 ff).

c) Scission and addition

Both resources are used supplementing each other in the extension of the mood and voice
systems.

The stems which in Anatolian (as the representative of Indo-European 2) function as
allomorphs of the indicative, that is, of anon-imperative, are classified in Indo-European 3 as
indicative stems and subjunctive stems: in both terms a process of allomorphization takes
place. I have explained elsewhere that there are no stems which are specifically of the in-
dicative or the subjunctive kind: The definition is proportional and this can be seen very
clearly when one touches upon the subject. By the time, however, the well-known fixed
correlatives of athematic indicative/thematic subjunctive or that of thematic indicative/
subjunctive with long vowel were established. Authors like Renou and Kuryfowicz by now

acknowledge at least the identity of the -e/o of the indicative and that of the subjunctive
(cf. Kurytowicz 1964: 137 ff; 1977: 90 ff.).

Thus a ternary system was formed. A positive term corresponding to the impressive func-
tion of the language (the imperative) had a negative one in which this function coexisted
with the representative one; then the above-mentioned split created a secondary opposi-
tion of indicative/subjunctive. However, on the other hand, there is addition of the opta-
tive, which has a characteristic suffix -ié/-i (also in the -oi- of the thematic verbs). This four-
term system is highly complex: the indicative indicates reality as against the subjunctive-
optative, and there is a gradation of reality (indicative)/eventuality (subjunctive)/possibility
(optative); at the same time, there is an opposition between mandate (imperative)/will
(subjunctive)/wish (optative) within the impressive values which these moods preserved, as
compared to the indicative.

One would achieve similar results by analysing the creation of a passive voice which makes
the binary voice system (active/middle) a ternary one; although this is not an Indo-Euro-
pean phenomenon since the different languages develop this independently, it is clear that
the old middle voice at times can be split into a middle and a passive one, whilst at others
the passive received the addition of special elements.

d) Combination

The evolution of the tense system in Indo-European 3 consists of two phases which are
chronologically different.

In the older of the two, the present/preterite opposition, which belonged to Proto-Indo-
European (Indo-European 1), takes a new formal characterization which does not cancel
out the older one (based on endings): certain stems are assigned to the present, others to
the preterite. The distinction is only proportional to a great extent and often dialectal;
however, as in other parallel cases, certain stems tend to be assigned to one of the opposed
terms, others to the other (cf. Adrados 1975: 675 ff.). The whole of this formal re-shaping
hardly affects the semantics of the opposition, except that one must assume that is was
formerly a privative one (-m, -s ...as the negative term) and that it is now an equipollent
one, for both the present and certain preterites in certain Indoeuropean languages display
atemporal uses.

The second phase is that in which, at least in Greek, Indo-Iranian and Baltic, the present/
preterite tense opposition was combined with the atemporal present (negative term)/
desiderative opposition, thus a positive future term being created. As is known, the syn-
thesis came about through the proximity of the semantic values of the desiderative and the
future and through the existence of a ‘case vide’ for the future.

Once more, we observe in all cases the secondary nature of the ternary and multiple oppo-
sitions in general and their construction through lexical elements, and above all stems.
There is only the exception of the dual, which was created on the basis of an opposition of
endings.

5. The new oppositions are still to be discussed, as I mentioned above: as I stated then,
they were all created on the basis of opposing stems. There are processes of allomorphiza-
tion of existing stems (and other new ones which were created) which contract binary
oppositions that later give ternary ones.

a) Scission (split)

I believe this is what occurs in verbal aspect. Leaving aside that based on the opposition of
base verb/verb with preverb as likewise the diverse ‘Aktionsarten’, whose existence in Indo-
European 2 I already mentioned, I shall keep to the aspect which comes into play between
the present, perfect and aorist stems. The present/perfect opposition starts from that be-
tween diverse allomorphic present stems and a deverbative of state.” In fact, the result is an
opposition between a base verb and a special stem with stative value.

But this base verb (the ensemble of allomorphized stems) splits into present and aorist by
means of a new process of allomorphization in both sectors. The starting-point is plausibly
in certain iterative and durative stems and in other punctual ones, which become crystalli-
zation points around which the two opposed terms are made up. Thus, the oldest and most
general binary opposition is extended to make up a ternary one (cf. Adrados 1974:675 ff.).

b) Combination

The degrees of comparison in the adjective (positive, comparative and superlative) make up
a ternary system resulting from the combination of two previous binary systems, the posi-
tive/comparative and the positive/superlative.

It is well-known that these are sometimes lexical oppositions, sometimes grammatical ones
(with diverse allomorphized suffixes), and sometimes mixed ones. It is clear that the first
step is allomorphization (as is known, for example, -tero and -ios originally had quite
different values) and the constitution of the two binary systems with a common term (the
positive). Their combination makes up a ternary system of the gradual type, as to a certain
extent that of the Indicative/Subjunctive/Optative (reality/eventuality/possibility) and that
of Imperative/Subjunctive/Optative (mandate/will/wish), although clearer.

As was stated, these new systems of Indo-European 3 display defective data, among others,
which definitely prove their secondary nature.

6. I think the foregoing is sufficient, although similar arguments may be put forward for
the ‘Aktionsarten’, deixis of the pronoun, etc., always along the same lines.

If Proto-Indo-European or Indo-European 1 was a non-inflexional language, as is thought,
an intellectual mechanism must be thought up to explain how later categories were created:
obviously with the aid of formal elements prior to these and originally indifferent to them.
The process took place in stages: the creation of the binary oppositions, the extension of
these latter, the later creation of other binary and ternary ones and their extension. Funda-
mentally, the procedures were always the same: if we follow up the process in the in-
dividual languages, we shall see how it is repeated. There is, however, an evolution: first the
lexical and ending resources are used, whereas the stems are hardly used; then, without
renouncing the former, the latter become the more frequent.

In all cases, abundant irregularities are left which may serve as a guide in the reconstruction
of the oldest oppositions and the development of the system through its various stages. I
have given a more detailed account of these irregularities and their use to this reconstruction
elsewhere (Adrados 1984b).

10
Notes

1 Cf. among others: Adrados 1974, Adrados 1975, Adrados 1981a: 96-122, Adrados 1982: 1-35,
Adrados 1985a und 1985b.

? Above all in Adrados 1985b.

3 Cf. Kuryłowicz 1956, and above all Kurytowicz 1964 and Kurytowicz 1977. Cf. a criticism indicating
the points which differentiate my work from that of this author in Kurytowicz 1974: 932 ff. and
Adrados 1982: 195.

Adrados 1962: 5-41, Adrados 1966: 131—154, Adrados 1968: 1—47.
That Anatolian had no masculine/feminine opposition is becoming acknowledged by all. Cf. lately
Laroche 1970: 5-—57.

6 Adrados 1974: 849 ff.; Adrados 1975: 7/9 ff., in which the supposed secondary loss of the sub-
junctive in Baltic and Slavic is also refuted.

7 For the origins of the perfect and its non-existence in Anatolian, cf. among others Nadia van Brock
1964: 119—165 and Adrados 1981b: 27--58.

References

Adrados, Francisco R., 1962. *Gramaticalización y desgramaticalización", in: Homenaje a André Martinet
3, edited by D. Catalan (University of Laguna Press), 5—41. Collected in: Estudios de Lingüfstica
General, 2nd edition, Barcelona 1974: 221—254 [Translated into German in: Sprache und Bedeu-
tung (München: Fink) 1977: 238—264].

Adrados, Francisco R., 1966. "Historische und strukturelle Methode in der indogermanischen Sprachwis-
senschaft", Kratylos 10: 131—154. [Translated into Spanish in: Estudios de Lingüistica General,
2nd ed. Barcelona 1974: 257-.283].

Adrados, Francisco R., 1968. "Die Rekonstruktion des indogermanischen und die strukturalistische
Sprachwissenschaft", /F 73: 1—41.

Adrados, Francisco R., 1974. Evolución y Estructura del Verbo Indoeuropeo, 2nd edition (Madrid:
Instituto ‘Antonio de Nebrija”).

Adrados, Francisco R., 1975. Lingüística Indoeuropea (Madrid: Gredos).

Adrados, Francisco R., 1981a. “Indo-Furopean s-stems and the Origins of Polythematic Verbal Inflec-
tion”, JF 86:96 .-122.

Adrados, Francisco R., 1981b. "Perfect, Middle Voice and Indoeuropean Verbal Endings", Emérita
49: 27 —58.

Adrados, Francisco R., 1982. “The Archaic Structure of Hittite: the Crux of the Problem”, JIES 10:
1-35.

Adrados, Francisco R., 1985a. “Ideas on the Typology of PIE”, JIES (in press).

Adrados, Franeisco R., 1985b. “Der Ursprung der grammatischen Kategorien des Indogermanischen”, in:
Acts of the VHI Fachtagung für Indogermanische und Allgemeine Sprachwissenschaft

Van Brock, Nadia, 1964. “Les themes verbaux a redoublement du hittite et le verbe indo-europeen”,
RHA 22: 119-165.

Fairbanks, C.H., 1977. “Case Inflection in Indo-Kuropean”, JIES 5: 101 -131.

Kurytowicz, Jerzy, 1956. L apophonie en Indo-Européen (Worclaw: Polskij Akademij Nauk).

Kurytowicz, Jerzy, 1964. The Inflectional Categories of Indo-European (Heidelberg: Winter).

Kurytowicz, Jerzy, 1977. Problemes de linguistique Indo-Européenne (Wroclaw: Polskij Akademij
Nauk).

Laroche, Emmanuel, 1970. ^Le probléme du féminin", RHA 28: 50—57.

Lehmann, Winfred P. 1958. “On earlier Stages of the Indoeuropean Nominal Inflection", Language
34: 179—202.

Rosenkranz. Bernhard, 1979. “Die Struktur der hethitischen Sprache”, ZDMG, suppl.

Schmalstieg, William R., 1980. Indo-European Linguistics; A New Synthesis (University Park and Lon-
don: Pennsylvania University Press).

Villar, F., 1974. Origen de la flexión nominal indo-europea (Madrid: Instituto “Antonio de Nebrija”).

Syllabic prominence in ancient Greek: a typological approach

William Sidney Allen (Trinity College, Cambridge)

It is almost universally accepted (disbelief being practically confined to modern Greek
speakers) that the accent of ancient Greek, up to about the third century a.D., was of
melodic type, manifested by a rise of pitch at a particular place in the word. The evidence
for this has been widely discussed: it includes the testimony of ancient writers, philosphical,
critical, and musical, supported by the implications of their terminology; of ancient Greek
musical settings, notably in certain of the Delphic hymns; and of parallelism with the
related Vedic accentuation, which was similarly described by Old Indian phoneticians.
There is also support of a typological nature if one accepts Trubetzkoy’s generalization
(1939: 213) that those languages whose prosodic phonology is most simply described in
terms of the ‘mora’ characteristically employ pitch as the accentual feature; for the rules
governing the location and form of the accent in ancient Greek are certainly most simply
stated by treating long vowels and diphthongs as comprising two units (morae), as against
the one of short vowels (cf. Jakobson 1962: 263; Allen 1973: 234 ff.).

It is further indicated that the melodic peak did not involve any appreciable increase of
intensity or duration such as may characterize a dynamic (stress) accent. For no account
is taken of the position of the accent in the composition of classical Greek verse; whereas
when later the accent changes to dynamic type, it does play an increasing and ultimately
exclusive ròle. One result of this is that readings of ancient Greek verse by modern Greek
speakers, which commonly stress the accented syllables and make no distinctions of
phonemic vowel-length, thereby destroy any element of regular rhythm: a line such as
Iliad 1.84, for example, of which the quantitative pattern is

—00—00—00—00—00—— (where —/o = heavy/light syllable),

being read as

[tón d'apamivó-menos prosé-fi pó-das ocí.s agiléfs] (where ' now indicates stress, and -
automatic vowel-lengthening in stressed open syllables),

in other words with a ‘rhythm’

'XXX'XXX'X'XX'Xx'

followed in the next two lines by x'x'Xx'xXx'xx'x'x and x "X'Xxx" "x'x"x.

Not quite the same agreement prevails regarding the typology of the classical Latin accent,
but the majority opinion is certainly that it was of dynamic type, in spite of the statements
of Latin grammarians based on Greek tradition. Some of those who appreciate the evidence
for a dynamic accent but are unwilling to dismiss the grammarians' evidence entirely have
had recourse to a compromise solution. They note that many descriptions of ‘stress’ in

modern languages suggest that it is not simply, or even principally, a matter of acoustic
amplitude or auditory loudness, but is commonly also associated with variation of duration
and particularly of pitch. Therefore, it is argued, we should not make a sharp distinction
between the two types of accentuation, since changes of pitch are common to both (cf.
Schmiel 1981: 22, with further references). This argument, however, obscures an important
issue. It is one thing for pitch to be the primary feature of an accent, and quite another for
it to be an incidental cue to some other physiological activity on the part of the speaker,
namely what we conventionally call ‘stress’. Precisely what is involved in the latter is much
disputed: one hypothesis that has found favour is that of Öhman (1967), who suggests that
stress involves “the addition of a quantum of physiological energy to the speech produc-
tion system as a whole... distributed (possibly unevenly) over the pulmonary, phonatory,
and articulatory channels”. The marked pitch component in stress is readily explained by
the fact that “the laryngeal musculature is often affected by tensions in the other muscula-
tures of speech” (Stetson 1951: 95; for more specific details cf. Lehiste 1970: 82, 125,
144).

As regards the distinction between the Latin and Greek accents, what the compromise solu-
tion overlooks is the lack of any evidence for cues of duration or amplitude in the ancient
Greek accent — just as in modern pitch-accented and tonal languages fundamental fre-
quency alone generally provides “the necessary and sufficient perceptual cue" (Gandour
1978). It also overlooks an important structural difference as shown by the rules governing
the location of the accent in the two languages. In Latin the position of the accent is
governed by the so-called ‘penultimate rule’: if the penultimate syllable is heavy it is stressed;
if it is not (in words of three or more syllables) the antepenultimate is stressed, whether
heavy or light — thus re-fé-cit, re-féc-tus, but dé-fi-cit, ré-fi-cit . It is to be noted that it
makes no difference whether the heavy quantity of the penultimate syllabie is ‘by nature’,
i.e. on account of its containing a long vowel or diphthong, or ‘by position’, i.e. on account
of its consonantal closure. In Greek, where the accent is relatively free, there are certain
limiting rules; but here the determining factor is not generally the quantity of the syllable
but the length of the vowel. A further difference from Latin is that in Greek a long vowel
or a diphthong may have two forms of accent, the acute as in ofkoi ‘at home’ or the cir-
cumflex as in oikoi ‘houses’, the high pitch in the former being located on the second mora
and in the latter on the first (equivalent to dikoi). No such distinction prevails in Latin
(pace Pompeius and Priscian), where the concept of the ‘mora’ is irrelevant (pace Trubetz-
koy and Jakobson: cf. Allen 1973: 161 f.).

Regardless, therefore, of the actual phonetic characteristics of the Latin accent in particu-
lar, it is clear that the accents of Latin and of Greek were quite different in terms of struc-
tural typology; and provided that we do not insist on the precise physiological implications
of their titles, we may continue to distinguish them in the traditional terms of ‘dynamic’
versus ‘melodic’.

It is however also the case that no languages, even those with melodic accents (or with
distinctive tonal systems), lack patterns of non-melodic ‘prominence’ (to use a non-com-
mittal term) — just as no languages, even those with dynamic accents, lack patterns of pitch
variation, although in both cases these secondary patterns may not necessarily be linked to
the accent: I have elsewhere (1973, 94 f.) listed a wide range of spoken languages which
illustrate this principle. Patterns of this kind are not marked in orthographies, since they
do not have the same differential functions as, say, the vowels and consonants, and so are

not directly traceable in the case of dead languages; and for the same reason contemporary
writers may make no reference to them even though they discuss accentual patterns which
may equally be unmarked. In the case of Latin we know nothing about non-accentual
patterns of pitch-variation, such as may have characterized the intonation of sentences; and
in the absence of any direct testimony it is difficult to see how we could set about recov-
ering them. But it is at least possible in principle that some indications could be recovered
of non-accentual patterns of prominence in Greek. For, as Latin poetry suggests, there may
be a tendency, even in quantitative verse, to seek coincidence of prominence (accentual in
the case of Latin) with the strong position of the foot, i.e. with that element of the foot
which is normally marked by a heavy syllable. If this were the case in Greek, then the
placement of words in the verse should give some indication of any (non-accentual) promi-
nence-patterning amongst their syllables.

Many words would of course be irrelevant to such an enquiry, since their quantitative
patterns would automatically restrict them to particular places in relation to strong posi-
tion, so that their placement would not necessarily indicate any element of prominence;
this applies to all words in which a heavy syllable is flanked (on either or both sides) by
one or two light syllables. But other types of word, containing only heavy syllables (or a
succession of more than two light syllables in the case of ‘resolved’ feet) could provide
relevant evidence, since their quantitative patterns permit alternative placements in relation
to strong position. And even in the case of the former type, the ‘tied’ word-patterns, it
would be suggestive if any of them were avoided in particular positions, or only admitted
‘faute de mieux’.

I have argued elsewhere (1973: 283 ff.) that an investigation of this type in spoken Greek
verse reveals certain very clear strategies of placement, which may be summarized (as I have
not previously done) in the following diagrammatic formula, showing how various types of
word (or word + clitic combinations) tend to be placed in relation to the strong positions
of the line (S), including those which for quantitative reasons must necessarily be so placed
if at all; optional syllables are set in parentheses:

>
00 O OO O

In deriving particular word-patterns from this formula by working from right (end of word)
to left, one may move (but not necessarily) from one row to the other, with the proviso
that restrictions on sequences of light syllables in verse preclude movement from (0) to o.
There is no spoken metre which could provide examples of word-pattems (S), _ S! al-
though they are derivable from the formula; and notable patterns nor derivable from the
formula are:

SIS ]J
00

S do sd e. lj S TS 1.8188) 8.0.8
000 00 00 o 00 00 0000 “000 000° op

These non-derivabilities reflect, inter alia, the constraints of ‘Porson’s Law’? in iambics and
trochaics and of ‘Gerhard’s Law’? and ‘Naeke’s Law”? in hexameters.

In interpreting the formula it is to be noted that it assumes the application of the tradi-
tional ‘law of indifference’, whereby at the end of a line of verse aheavy syllable in terminal
weak position is treated as light, and a light syllable in terminal strong position is treated as

heavy. Thus a hexameter may end with a word(-end) _ _ (placed S_) because it is equiva-
lent to _ (= catalectic dactyl), and an iambic trimeter or catalectic trochaic tetrameter
may end with a word(-end) ba (placed op because it is equivalent to . When this is
taken into account, the derivations from the formula agree with the data in 90—95% of

cases, and virtually 100% in the coda of any line.

The exceptions concern primarily placements of the type (. . . à . The most notable of
these is the position preceding the penthemimeral caesura of iambics, where there may be
special considerations. There are also the exceptions to ‘Naeke’s Law’ in Homer (as against
Callimachus), involving principally the placement of words or complexes of pattern — —
ending with the 4th foot, where however there is clear evidence of a ‘faute de mieux’
principle at work (cf. Allen 1973: 286 ff.). Other rather frequent exceptions concern the
beginnings of lines (e.g. placements a È in iambics,, S_ in hexameters); but rhythmical
irregularities and inversions are here a general commonplace. Another less common ex-
ception is the placement S È in iambics, but here there are special constraints on the
nature of the final two syllables (cf. Allen 1973: 323 f.).

The fact remains that there is a very clear preference for particular placements of words
relatively to the strong positions of the verse line, as summarized in the above formula; and
it seems most likely that these preferences reflect certain patterns of prominence in Greek
words (unrelated to the accent), and more specifically that those elements which are prefer-
ably placed in strong position were in some sense prominent in the language. From the
placement formula we could then derive the following rules for the incidence of prominence
in Greek words:

1. Prominence applies to an element constituted by either (a) one heavy syllable or (b) two
light syllables.

2. Words (or word-like sequences) longer than an element have internal contrasts of promi-
nence/non-prominence.

3. If the final syllable of a word is heavy, it is prominent.

4. If the final syllable is light, the next preceding element is prominent.

5. A preceding element separated? from the prominent element is also (secondarily) promi-
nent.

In earlier discussions I have suggested that ‘prominence’ is here to be equated with stress —
though presumably weaker than the accentual stress of a language like Latin (as is also
suggested by its effacement by the accentual stress which later replaced the melodic
accent). Some scholars have been convinced by these arguments; others, however, whilst
accepting that the preferences of placement indicate some form of prominence in partic-
ular syllables of words, would contest the conclusion that this was stress: enhanced dura-
tion or raised pitch have both been proposed, but not as cues to stress. I have briefly
mentioned (1973: 292) some objections to these counterproposals; but there must in-
evitably remain an element of subjective judgement in assessing probabilities unless more
positive clues can be uncovered.

We might consider the patterns of assumed prominence expressed in the Greek formula as
a kind of typological ‘fingerprint’ of that type of prominence. And if it were possible to
find a matching fingerprint in a language where the phonetic nature of the prominence was
established, we should have quite a strong typological argument for concluding that the
prominence in Greek was of the same nature. And it is at this point that we return to the
question of the Latin accentual system.

The traditional ‘penultimate rule’ of Latin, whilst descriptively adequate, has certain weak-
nesses in respect of explanatory adequacy (in the sense of Chomsky 1965: 27). For it
implies that, whereas an antepenultimate light syllable is accentable under the same condi-
tions as a heavy, viz. if followed by a light syllable, a penultimate light syllable may not be
accented (except in disyllabic words). It might be and often has been suggested that the
accentuation of an antepenultimate light syllable is simply a ‘pis aller” in the absence of any
heavy syllable in the accentable portion of the word. But words of this accentual pattern
are very common, being no more constrained than any other, and it is explanatorily un-
satisfactory to treat them simply as exceptions.

Quite independently of the Greek question I therefore attempted some time ago (1969) to
formulate a more satisfactory rule from this point of view. It involved the concept of an
accentual ‘matrix’, which comprised either one heavy or two light syllables; phonetically
the peak of stress in a disyllabic matrix would be reached on the first syllable, just as it is
reached in the first part of a heavy syllable, following the commonly ‘diminuendo’ pattern
of stress. We could therefore represent the accented matrices as ^ or 6 à respectively. In
the latter the first of the two syllables (carrying the stress-peak) would correspond to the
syllable traditionally described as accented. On this basis the location of the Latin accent
(in words of three or more syllables) may be stated as follows:

The accent occupies the last matrix in the word preceding the final syllable.

The last part of this rule excludes accentual patterns such as o — “or — ó ò. As stated in
these terms there is nothing peculiar or exceptional in the accentuation of e.g. ré-fi-cit,
which is now analysed as ó ó 5, accented on the matrix o o, just ase.g. re- fé-cit is analysable

a$0 2g.

The rule for the location of the Latin accent, when thus stated, may be represented as
follows (where A = accent):

(ISOS

The rules of derivation from this formula are the same as stated for the Greek S-formula
above, Geer word-end accentuations

åo- o^o ; La o (in traditional terms ~0 5,005 2 sch

O 3
The uu pattern À _ o Will also apply to su words of this shape (thus e.g. ramus
as amämus).

It will be seen that this formula bears a remarkable similarity to the end of the Greek S-
formula. The only difference is accounted for by the irrelevance of final quantity in the
Latin accentual system. Other languages having a similar system to that of Latin are Sanskrit
(after the loss of the melodic accent), the modern Indo-Aryan languages, and classical
Arabic; and in the last two of these final accentuation does in fact occur in the case of
overweight syllables (-VC or VCC). Latin does have a few cases of final accentuation of
such syllables as the historical consequence of contraction, syncope, or apocope (e.g.
fümät < fümauit; nosträs < nosträtis; illic, illine < illice, illince), but there is no syn-
chronic rule to this effect? (thus e.g. laudas, laudant).

It is generally assumed that in longer Latin words there were also secondary accentual
patterns in the pre-(primary-)accentual portion (see e.g. Fraenkel 1928: 351 f.). If we
assume that the accentual rules represent in some sense a limitation set on the span of

prosodically undifferentiated sequences, an a priori hypothesis would be that the pre-
primary-accentual part of a word was treated as a word for secondary accentual purposes.
And in fact for classical Arabic the following rule is stated by Abdo (1969: 73a):

“Starting from the segment immediately preceding the stressed vowel, apply the rule once

more to any segments that again meet its structural description. The vowel stressed in the
second application receives secondary stress”.

Thus e.g. kätabäthu, zivàratuhunna, and with tertiary stress wäyatakätabüna. Metrical
evidence at least does not contradict this assumption for Latin (e.g. in iambics/trochaics
sáspicábar, indiligentia, uinösissima, misericórdia; and at hexameter-end Ldodamia, dbscon-
dántur); and if we extend the Latin formula to include secondary accentuation, the similar-
ity to the Greek formula becomes the more striking: thus:

50 () 5 49 00) 5 beside jj (o) 5 5 Oo
With the one exception noted, the formulae are identical.

The Greek placement rules gave rise to the hypothesis of some factor of prominence in
those elements of words placed in the S position. Typologically therefore it is at least a
reasonable conclusion that the elements of Greek correspond to the matrices of Latin, and
that S corresponds to A. The factor of prominence involved in A is generally agreed to be
stress, and it would be a rather remarkable coincidence if the prominence factor associated
with S were of a totally different kind. The exact correspondence of the disyllabic stress
matrices and disyllabic strong positions may then also suggest a phonetic explanation of
the phenomenon of resolution.

By contrast, the limiting formula for the recession of the melodic accent of Greek would
bear little typological resemblance to the A/S formulae. As we have seen, it is based on
vowels rather than syllables; it involves inter alia the concept of the ‘mora’, and (for its
most economical statement: cf. Allen 1973: 234 ff.) of a ‘contonation’ comprising either
the two morae of a long vowel or a diphthong (= the marked circumflex) or else the second
or only mora of the vowel plus the next vowel in the word (= the marked acute): the rule
is then statable that ‘not more than one mora may follow the contonation’: thus e.g.’
lupou, dööron, mönön, höloös, bainoò, dnèmos, dnthroôpos are admissible, but not e.g.
*doorois or *dnthroopois (correct are doörois, anthroöpois).

The one disagreement between the Greek (S) and Latin (A) formulae also deserves some
further notice. In the Latin formula the quantity of the final syllable is indifferent, but not
so in Greek; consequently in Latin the final syllable is non-A whether heavy or light, but in
Greek it is non-S only if it is light. This means that in Greek, unlike Latin, there is the
possibility of alternative placements of certain words, viz. those ending in VC, depending
on the initial of the following word: thus e.g. the word neas in

S S.
11. 1.306 neas eisas

d S
but 493 va te propasas

which suggests the possibility of shifting prominence according to environment. This is
something which does not occur in languages like Arabic and Indo-Aryan, which permit
final accentuation, since only ‘overweight’ syllables are thus accented, and these by their
nature are not subject to syntagmatic variation of quantity (the same applies to the few

cases of final accentuation in Latin). But the crucial point here is that in Latin and these
other languages we are dealing with fixity of accentual placement? , whereas the assumed
Greek prominence patterns are non-accentual, and it would not be surprising if syntag-
matic variation in these were found: rather as in a dynamically accented language like
English words may bear different melodic (intonational) patterns according to their en-
vironment. In other words, in Latin and similarly accented languages we are dealing with
specifically word-patterns of stress (just as with the pitches of the Greek melodic accent),
whereas in Greek we are dealing with sentence-patterns of prominence, albeit regulated by
word-boundaries.

Conclusion:

The rules governing the melodic accentuation of ancient Greek are quite different from
those governing the dynamic accentuation of Latin. But the rules relating the strong posi-
tions in spoken Greek verse to the quantitative patterns of Greek words are almost identical
with those relating the Latin accent to the quantitative patterns of Latin words. This
detailed coincidence in independent prosodic systems suggests that the physiological factor
underlying both sets of rules is the same. If in the case of Latin this was stress, it is reason-
able to think that the same factor underlies the Greek rules, though in this case it has no
accentual function.

Notes

The pattern o — 8 in iambics and trochaics is precluded by rules constraining word-division in reso-
lution.

2 R.P., Hecuba? 1802, Suppl. ad praef., xxx.

3 E.G., Lectiones Apollonianae, 1816: 142 f.

4 A.F.N., RAM 3, 1835: 516 f.

5 Le. by a heavy or by one or two light syllables, of which the first and last, though themselves ele-
ments, are not separated from the (primarily) prominent element.

$ For further discussion cf. Allen 1983.

7 To conform with the moric basis of the rules long vowels are indicated by double writing. The coda
of the disyllabic intonation may be assumed to occupy the whole of the vowel/diphthong involved,
ie. to return to the non-high pitch level towards the end of it (as in the case of the monosyllabic
contonation); the grave sign is therefore marked on the second element of the long vowel or diph-
thong.

8 It might even be argued (cf. Allen 1983) that the limitation of final stress to overweight syllables has
the specific function of ensuring such fixity.

References

Abdo, D. A., 1969. Stress and Arabic Phonology (Dissertation Ph.D. Illinois). Published as: On Stress
and Arabic Phonology: a generative approach. (Beirut).

Allen, W. Sidney, 1969. “The Latin Accent: a restatement", JL 5: 193—203.

Alien, W. Sidney, 1973. Accent and Rhythm (London: Cambridge University Press).

Allen, W. Sidney, 1983. ‘Some Reflexions on the Penultimate Accent”, Illinois Classical Studies 8 (In
honor of John L. Heller).

Chomsky, Noam, 1965. Aspects of the Theory of Syntax (Cambridge, Mass. M.I.T. Press).

Fraenkel, Ernst, 1928. /ktus und Akzent im lateinischen Sprechvers (Berlin: Weidmann).

Fromkin, Victoria Alexandria (ed.), 1978. Tone: a linguistic survey (New York: Academic Press).

Gandour, J. T., 1978. “The perception of tone”, in: Tone a linguistic survey, edited by V.A. Fromkin
(New York: Academic Press).

Grotjahn, Rüdiger (ed.), 1981, Hexameter Studies {= Quantitative Linguistics 11} (Bochum: Brock-
meyer).

Jakobson, Roman, 1962. “On ancient Greek prosody”, in: Selected Wrirings 1 (The Hague: Moutor:),
262-271.

Lehiste, Ilse, 1970. Suprasegmentals (Cambridge, Mass.: M.I.T.-Press).

Öhmann, Suzanne E.G., 1967. “Word and Sentence Intonation: a quantitative model", in: Quarterly
Progress & Status Rep. 2-3, Royal Institute of Technics (Stockholm), 20.

Schmiel, R., 1981. “Rhythm and Accent: Texture in Greek Epic Poetry”, in: Hexameter Studies, edited
by R. Grotjahn, (Bochum: Brockmeyer), 1 ff.

Stetson, R.H., 1951. Motor Phonetics (Amsterdam: North Holland Publishing Company).

Trubetzkoy, Nikolaj S., 1939. Grundzüge der Phonologie (= TCLP 7) (Prague).
(2nd edition Göttingen: Vandenhoeck & Ruprecht, 1958).

Hittite pai-, why no *wai-: an Anatolian-Indo-European heterogloss!

Yoél L. Arbeitman (Bronx, New York)

This little article honors my one-time
professor whom I neglected as a student,
giving his Greek Dialects class but the
preparation time that Akkadian and Sans-
krit left (which meant c. 30 minutes from
an 18 hour study day).

Those of us who consider Anatolian as equivalent in worth to all of Indo-European Proper
(IEP) together for the recovery of the earliest parent language have received our share of
sneers. The view, which waxes and wanes in its intensity, that Anatolian ‘discrepancies’
vis-à-vis IEP are to be ascribed to Anatolian’s being a pidginized IE (sic!), from whose in-
formation bits we can explain away embarrassments for Brugmannian IE? by declaring the
item in question to be sub-/ad-/superstratic material, dominates amongst those who hold
the major seats on ‘The Board of IE Trustees’. In the allotted space I cannot approach this
grand question, which — at all events — is more one of authority than of scientific quest. It
has even proven impossible to deal with the three Anatolian lexical bits, whose cohesion
formed the foundation into which the single bit to be discussed herein, fits. Given these
realities, I will devote this space to my third utterance? on the Anatolian verb ‘to give’
(pai-[piya-).

I deem it worthy of first place to note that, in either camp described above, one ineluctable
actuality pertains for any Anatolian lexical bit: a choice which so far has been conceptual-
ized as only tripartite:

(a) a word or morpheme is obviously common to Anatolian and IEP;

(b) the Anatolian word is obviously borrowed from Hattic, Hurrian, Semitic, or unattested
substrat languages (Kultur- und Wanderwort often);

The model, that the evidence accumulated for this paper (but not spatially **presentable")
has led to, holds that there is a fourth alternative, one which sadly has been underutilized;

(d) Anatolian may present the lexical bit in an appearence sufficiently different from its
IEP cousins, that the family resemblance goes unrecognized? . When this choice is taken,
the equation can only stand when its vitality (1)is not vitiated by internal Anatolian
patterning and (2) is supported by parallel Anatolian: IEP heteroglosses.

20
1. History of the Proposals

In my article (1976: 79—80) I offered a triad of reasons that renders the now canonical
analysis of pai-/piya- into the constituent elements, preverb pe-° + Vai-, untenable. One
— not the least — of my three causes for dubiety was the absence of (a) a directionally
contrasting verb *wäi-/wiya- ‘receive’ and/or (b) a simplex *ai-/iya- ‘exchange’, in contrast
to pai- ‘go’, where (1) Hittite has (a) the contrast paizzi/wizzi (pe-/u- + Vei-), (b) remnants
of the simplex (e.g. the imp. i-t); (2) Cuneiform Luwian has an i-/awi- contrast (see the
references in my article [1974: 71—72]); (3) Hieroglyphic Luwian has the simplex (e.g. the
[rhotacized] 3rd sg. form iri goes") and the form prefixed by the frozen adverb aw- (3sg.
aw-i-ti ‘comes’). As I saw the situation, one of the signs for securing a definitive etymolog-
ical analysis of an Anatolian verb as belonging to the class of [root + preceding frozen
adverb, with Proto-Anatolian univerbation] is minimally the presence of such a bi- (ideally
tri-) partite contrast (simplex/pe-/u- forms). Unbeknownst to me, previous to my estab-
lishing these criteria, a storm had already been raging concerning the putative extancy of
the simplex i- ‘give’ in a Hittite nominal derivative. Had this much-discussed item even-
tuated to have any substance, I would withdraw the entirety of my insistence that pai-
‘give’ is a simplex and no compound. Carruba (1966: 16, fn. 17) — comparing another
item -- declared:

“Für die Bildung denkt man an die nachträgliche u-Thematisierung eines Verbalnomens auf
-uar, wie etwa in iuaru- “Gabe, Mitgift”, aus einem im Heth. nur noch in Komposition be-
legten *ai-/ila- “geben”, etwa in pai- gl. Bdtg.” (emphasis mine).

Weitenberg (Anatolica 4. 1971/1972: 166— 167) goes one step beyond:

“Dieses Verbum [*ai-/iya-] ist nach allgemeiner Ansicht das Simplex des Verbums pai- “ge-
ben", das zusammengestellt ist aus dem Prüverbium pé- und eben diesem *ai-. (. . .) Diese
Etymologie ist attraktiv, (. ..) Es sei aber daraufhingewiesen, daß es andere Etymologien
van [sic] pai- gibt: Verwandtschaft mit Toch. B pito "Verkauf" und ahd. feili “verkäuflich”
wird vorgeschlagen (Cop... .) und Rosenkranz (. . .) erklürt es zu ep-/ap- “nehmen”.”?7

Puhvel (19812: 213) notes re Carruba's other item: “(It) is derivationally strange and not
much helped by Carruba’s parallel iwaru- ‘gift’ (alleged *i-war, verbal noun from the root
of p-ai-, p-iya- ‘give’)”. It would have helped had Puhvel provided the basis for his “alleged
* *" by the simple means of a reference to his 1958 article (reprinted Puhvel 1981b: 46,
fn. 2). What Puhvel had concurred with and Weitenberg had rejected was Speiser's (JAOS
53. 1935: 436 fn. 17) loan derivation of the Hittite word from Hurrian. Speiser's deriva-
tion is surely correct (the Hurrian word-family he associated iwaru- with is discussed at
length in my article [1982: 20])5. Thus, piya- ‘gives’ remains isolated within Anatolian,
with almost no basis for being analyzed as p-iya-. An examination of the history of the
canonization of this segmentation will provide the contextual perspective enabling us to
divest ourselves of this enshrined item, the challenging of which verges on heresy.

The earliest etymological speculation is represented by the perplexity of Hrozny (1917: 14,
fn. 6):

“Dieses sehr häufige hethitische Verbum (3.P. Sg. Präs. pa-a-i, pa-iz-zi), das m.E. etwa die
Bedeutungen “ziehen, gehen, fahren; geben, zahlen; nehmen(?)” hat, klingt an ai. pati “hü-
tet, schützt" gr. *náouac "erwerben" an. Wegen der Verschiedenheit der Bedeutungen ist
indes ein Zusammenhang sehr fraglich."

The comments (1917: 163) are of great interest and in the Glossar (235, fn.1) he has dis-
tinguished päi ‘er gibt’ from paizzi ‘er geht’. One must needs admire the frank use of the
word ‘anklingen’ vs. (etymological) ‘Zusammenhang’. We were still ‘pre-preverbial’. The
dominance of the preverbial era for this verb begins with Petersen (1933: 32), whose pro-
posal was couched in circumspect word-choice:

“Further evidence of the close relationship of Hittite and Tocharian is the existence in
both languages of two common roots which are not found (at least in the same sense) in
the rest of the IE territory.? ... The second of these roots is *ai- ‘give’, found most clearly
in Toch. e-, e.g. e-st ‘thou gavest’ . . . In Hittite this *ai- probably appears in pe- pai- ‘give’,
e.g. pe-hhi ‘I give’, ... inasmuch as the prefix pe- ... was contracted with the root. The
development of this root evidently goes hand in hand with the loss in both Hittite and
Tocharian of the IE *da- ‘give’, which could not maintain itself because in both languages
it had become indistinguishable from *dha- ‘place’, and it was only the latter (in the sense
‘take’ in Hitt.) that continued to exist.” 19

Pedersen (1938: 115), after proffering the Tocharian-Hittite connection (Götze/Pedersen
1934 with references to earlier works which I have not seen), shortly changed the circum-
spect ‘probably’ of his namesake [-voice in the dental] to a definitive ‘gewiss’:

“Eine weitere diphthongische Wurzel steckt in pi-ih-hi ‘ich gebe’ . . . Gewiss (WALTER PE-
TERSEN Lg. IX 32) zu tochar. B ai-tsi ‘geben’ mit dem vor einem Konsonanten pře ge-
schriebenen Präverbium 'hin-’; pr-ih-hi also aus *pe-ehhi.”

Abros &pa and in the subsequent 45 years few have been those who doubted!!.

2. The Genesis of a Fresh Proposal and the -X Factor

The second part of the last sentence of my 1976 article concluded: “... else we must
throw overboard the idea that piya- is acompound.”. Thus, in clearing away the rubble of
a wobbling building, I raised no new structure to occupy a now vacant lot. Shortly there-
after Warren Cowgill wrote me!’ :

“I have for some time myself believed that these words are not compounds of a preverb
‘away’ with a root cognate with the Tocharian words for ‘give’. One of my reasons is the
same as one of yours: the absence of a correlate form with a preverb meaning ‘hither’...
The other is that Anatolian pai-/pi- shows exactly the same vocalism as dai-/ti- ‘place’,
which I am fairly sure comes — directly or indirectly — from *dhox-/dhx-, the o-grade and
zero-grade respectively of *dhex ‘put’. This would strongly suggest that the word for ‘give’
comes from a root *pex- or *bhex-. But the preverb Hitt. pi-e in my opinion comes from
a form (pay or *bay, which monophthongized to pe or be in Common Anatolian when
word- or syllable final, but remained with its original vocalism before vowel, as in paanzi
from *payanti * *pay yenti (with the two y’s coalescing into one, it seems). And a Tocha-
rian root ay- ought to imply an initial laryngeal, presumably my *x, which ought to appear
in Hittite prevocalically as (-h)h-, so that if the communis opinio about Hitt. pai- were
correct, the form ought to be rather *pehhai [by “the form” here, Prof. Cowgill obviously
intends the form of the Hittite root, not to be confused with pehhi (1st sg.), where the
laryngeal represents the desinence!]. To explain the attested shapes, we'd have to suppose
that the root instead had Kurylowicz's 84, so that *payai- could merge to *pai-, and then
with loss of distinctive vowel length a new zero-grade be created on the model of dai-/ti-
etc. Your point about reduplication being anomalous if the form is really an old compound
I had not considered, and is good."

Cowgill's neat equation of the root configurations — within Hittite -- of ‘place’ and ‘give’
forms the basis for the present proposal. One would like to infer from his analysis that he
considers ‘give’ inherited Indo-Hittite material, but the determination of the Indo-Hittite-
ness or Indo-Europeanness of a given item, in an Anatolian language or in any single Indo-
European Proper language, demands — minimally — something of cognateness in another
Indo-Hittite daughter or niece language. I believe that a cognate of Anatolian pai- is present
in Indo-European Proper, but in raiment so disguised that (1) our Brugmannian glasses fail
to perceive it and (2) the employment of the 'fourth alternative (d)' elucidated on p. 19
above, places it readily at hand. The development of Anatolian ‘place’ may reconstructed
as follows:

o-grade: 3.sg. *dPoEYe > *doye > *daii > dài

(-grade: 3.pl. *d'EYonti 2 *dyonti > *diyanti > diyanzi, parallel to which we have the
verb ‘give’:

o-grade: 3.sg. *bl'oEYe > *boye > *baii > bai

(-grade: 3.pl. *bhEYonti 7 *byonti > *biyanti > biyanzi” (cf. on the pre-Hittite recon-
structions Puhvel [1960: 55—56] and Kronasser [1956: 165—66 and 197], each of whom
conceptualized the glide in "they place" differently, for obvious reasons; my own con-
ceptualization approximates Puhvel's far more than it does Kronasser's).

The Indo-Hittite root *dheEY-/dhEY- shares a peculiarity with the Indo-Hittite root
*deAV"/dAV^ (alias *deO-/dO-) and with the roots *steA-/stA- and *yeE-/yE- in that all
four show -k aorists in Greek (in the active singular). Beyond this, the first and last of these
four have this -k incorporated into the root in Latin (and sporadically elsewhere — see
below). In Anatolian there is no trace of this -k either as a synchronically segmentable
morpheme or (even less) as meshed with the root.

The appearence of this -K in the Greek aorists édnxa, ééwka, éornka, fka and the root-
incorporation of this 'velar 'stem-forming' suffix' in Lat. facio, iacio are not to be viewed
as dialectal peculiarities. Rather these forms are still visible stages of a continuing process
of a major recasting of roots, on-going throughout Indo-European Proper, which (1) would
eventually result in neo-roots which are not segmentable into their constituent elements,
the second of which is this velar extension, and (2) forms a major heterogloss between the
Anatolian continuant of Indo-Hittite and Indo-European Proper. Corollary to this is the
heuristic tool, thus gained, of prima facie scepticism of proposals for Anatolian verbs which
include such extension (see section 3).

In my article (1981: 959 ff.) I dwelled on the numerous Indo-European Proper neo-roots,
constructed of Indo-Hittite Y mei- * extensions -k/£/gW" (and -L317). The sole datum from
that study relevant here is that, in the Indo-European Proper daughter languages, (1) the
use of the neo-roots predominates by a large measure over the use of the basic, unextended
root and (2) the velar extension apprears in the manifestations [ tvoice, +labial com-
ponent]. In what follows I shall employ the archiphoneme symbol -K to denote this
unitary set of extensions of ‘velar’ nature | tvoice, +‘palatalization’, +labialization]. The
operative model, while founded on the above facts, in no way denies (1) the wide use of
other extensions in Indo-European Proper (whether still segmentable or root-incorporated)
or (2) the continued survival and use of the unextended ‘proto-root’ in the Indo-European
Proper dialects (differing from dialect to dialect). What the model does demand is simply
that (1) the large-scale employment of the neo-roots is an Indo-European Proper innova-

tion and (2) the extensions that coalesce with the proto-root are true consonantal ‘suffixes’,
not laryngeal reflexes or such.14

This on-going process in Indo-European Proper can be divided into two major divisions:
(1) the one-time extension is already root-incorporated in Indo-European Proper ‘proto-
times’, i.e. for Proto-Indo-European Proper as a whole [old root + extension] already is
become [(neo-)root]; (2) the development is accessible, to the surveyor of the Indo-Euro-
pean Proper panorama, as occurring or having occurred either (a) monodialectally or (b) in
dialect-clusters. As my prime exhibit, I cite the chestnuts of the Latin, etc. reflexes of the
universally accepted neo-roots *d'e/ak- and *ye/ok-:

Latin fac-io (< *dPak- + yo) feci (< *dhek-)

Umbrian facia (pres. subj. 3sg.) feia (< *dhe[k?]-)15

Oscan fakiiad (iy. en pig

Venetic vhax-s-20!6 (pret. 3rd sg.)

Phrygian (a8 -) Baker (pres. 3sg. them., 'afficit")/é6aec (pret. 3rd sg.).
The same zero grade of the neo-root, which is characterized with a *-yo-/-i- suffix in the
present of Latin, etc., — marked with an -s — serves as the preterite base in Venetic. In

Phrygian the -k extended form is apparently specialized for the present, making the ap-
parent similarity of ‘sigmoid’ preterites in Phrygian and Venetic of minor significance. No
cognates of iacio are known from the geographically Italic languages (or from Phrygian).

We can speak of *dhék- as a neo-root in a cluster of Indo-European Proper dialects inasmuch
as (1) the position of Venetic as genetically Italic is less than secure!” and (2) the relation-
ship of even Oscan and Umbrian to Latin as a close-knit Indo-European branch (as e.g.
Slavic) is less than one would desire.!8

Thus, the question of root-incorporation, as a completed phenomenon, remains open even
within so-called “Italic”. Besides the Umbrian controversy, Venetic has nothing to compare
with the form we have. Other Indo-European Proper languages have stems of *dhe- and
*ye- with this -k extension, but — where we see it — it functions, as we may dare conclude
it does in Phrygian (to the extent that we venture to utilize Phrygian forms at all, esp.
considering the c.millennium that separates the cited forms). The synchronically non-
syrrhizoid Lat. -do compounds are another evidence-datum of original non-root-incorpora-
tion for the diachronist only.'?

In the case of *dhe-,it is a recurring nightmare of mine to imagine what if we had only
these Latin, Oscan, Umbrian, Venetic, and Phrygian + [-K] forms with which to reconstruct
the root for the proto-language. We would naturally assume an IE (or even worse, IH)
Ydhek- or more likely Y'dhak-. I always wake up from that nightmare to the reality of Gk.
Tiènui, Skt. didhati and most of all Hittite dai, whereupon I recite the incumbent prayer
"Happy is the linguist whose data are not limited." Yet the process could have been in-
exorable; we could have had

Gk. *réSexue (**7Waxpt) | or, indeed, *9écow (**9dc0w) <v*dhak- + -ye/o- et sim., with
tieku (**iaxpu) *iéoow (**iaoow) presential suffix in place of
presential reduplication

Irrespective of my own oneiratic proclivities, this never did come to pass across Indo-
European Proper in the so-far cited verbs; any budding linguist, who progresses from Latin

to first year Greek, recognizes that the -k is not radical and may even be eager — when he
compares the anomalous velar in these four Greek aorists with the -x of the perfect — to
extract a ‘‘past’ function’ as being inherent in ‘the morpheme’. So we will have to call a
more convincing witness, one whose -K has come closer to root-incorporation across the
dialect spectrum, one who has experienced the process almost being carried forth to its
logical consequence, but not quite: VY yeu-g-, best rendered by Germ. ‘fügen’. As a verbal
root — in its prime meanings”? -- it survives unextended only in Indolranian (e.g. Skt.
yaüti besides yundkti?! ). We can speak here, quite properly, of a verb almost lost to non-
extendedness; the linguist who, when he progressed to elementary Greek, recognized the
non-radical nature of the velar in facio/feci: ri9nuu/é8mka, now has both to untertake
Sanskrit and become more of a scientific segmentator. It is with a good deal of justice that
Pisani (1974) gives the roots as dhe, dhek-; ie, jek; and I. ieu, II. jeug, respectively.

We have sketched so far examples where (1) the -K as root-incorporated was limited to a
single language or small cluster of languages and its origin as an extension could be readily
detected by the addition of multitudinous cognate data (the case of facio and iacio) and
(2) almost all of the data of almost all of the daughter languages would present the -K as
originally radical except for the addition of the archaic remnant material, limited to one
group (the case of Yveu- +[-g-]). The comparative method must be built on extant material,
but — as a science — is required not to stop with this material; it must and does extrapulate
principles and patterns from the data(particularly remnant archaisms) that have been (ac-
cidentally) left. It is natural that there are cases of (3): where only the application of the
methods, thus gleaned, can recover original relationships in the absence of any extancy of
relationships having the neatness of examples of types (1) and (2) above. I have chosen, as
the exhibit for this kind of less puissantly apparent relational extancy, a root of the same
semantic domaine as the prime topic of this paper. For purposes of summary citation, the
forms and definitions are reproduced from Pokorny. He, however, presents what is here
conceived as root and [proto-root + extension] having become neo-root, as two separate
items with no nexus. Pokorny’s 3. sel- ‘nehmen, ergreifen’: Gk. &Aeiv 'nehmen', Awp
“Beute, Fang’; (base slefi)- cf. Watkins [1962: 179]), O.Ir. adroilliu ‘verdiene’ (ad-ro-sli-);
O. Icel. selia, OE sellan ‘übergeben, verkaufen’; P’s släg*- “fassen, ergreifen’: Gk. Hom.
&AAafie (*éoXafe) nehmen, fassen, ergreifen’, Aegin. Maßwv, epic. Ion. Adfouaı (Mar u).
tw), “nach aivvpaı ist ion. att. Adlunaı, boot. \aS5ovoÿn umgebildet”; OE laccan “fassen,
ergreifen’ (= Aafouaı aus *slagtiö), NE latch””.

The neo-root(!) *slagW- is built of proto-root *sel- in theme II: *s/- * (full grade, a variety
of) labialized -K, as seen e.g. in the neo-root *meig* (theme I + -K). It eventuates that the
extancy of apparent relationship is not that much less in (3) than in (1) or (2) and our lin-
guist, who learned Greek and recognized -K in facio, progressed to Indo-Iranian and un-
covered -K in Vyeu- +/-g-), should have detected -K within the Gk. &Xew: *gAaßew relation-
ship itself.

All three above cases have -K in one or more Indo-European Proper dialect, but not in
Common Proto-Indo-European Proper.

I stated earlier (p. 22) that ‘alternative d’ will unfog our Brugmannian glasses, revealing the
IEP cognate of Anat. pai-/piya-. In this Indo-Hittite: Indo-European Proper connection, I
propose that we have the following relationships:

IEP neo-roots

Theme I. + -K- Theme II + dK-
Vane. (< *dheEY-) +/-k) veel +(-dgW°)
Yyeu- *(-g-) V bhag- (Y bh[e EY +[-dg-]).

It is but one variable that divides y b''ag- from the three other exhibits: the segmentation is
not perceptible within Indo-European Proper, i.e. the root-incorporation of -K is already
complete in Proto-Indo-European Proper and our linguist has had to go beyond 'Indo-
Greek' to Anatolia. The ‘economy’ with which the etymology is proposed here is a factor
of spatial constraints. The phonological and laryngealogical considerations do, however,
perform their constituent roles in the proposal. An overview of the relationships between
*dhe. and *bNag- may be given in a proportion, combining Cowgill’s original terminology
and my own: Proto-Indo-European Proper neo-root *bNag- ( *bhx-dg-): IH *bheX- = “Latin”
neo-root *dhek- {*dhét-k-): IE (and IH) *d'ex..

Much as IH morphemic relics continue sporadically in Indo-European Proper (e.g. Y yeu- ^v
Vveug-), so too do semantic archaisms make their way, on occasion, through the ‘Indo-
Hittite: Indo-European Proper barrier'. As well-known, the semantic opposition partner of
Hittite päi- is a cognate of Indo-European Proper Vdo-, a root which in Hittite has the old
meaning ‘take’ vs. the Indo-European Proper meaning ‘give’. We have seen that in Greek this is
one of those four roots which show the initial stages of a -K extended root (aor. €5wxa).
Just as in facio and iacio Latin has completed the root incorporation, I believe that one dia-
lect of Indo-European Proper has a neo-root, built of *do + [-K root-incorporated], the ety-
mology of which would long ago have been recognized if the semantic variable, fed into
etymologizing, did not exclude Anatolian ‘divergencies’. This monodialectal Indo-European
Proper neo-root *d0g- (< *deAW-g-, matching Gk. ewka, but with [+ voice -K]) is the
etymon of ON pret. tok, while the ¢-grade *dag- (< *dA™g-, formally equivalent in stru-
ture to the first syllable of Lat. facio) yields the ON pres. taka, which are used today in
English as fook, take respectively. The model of (1) preservation of archaic meaning and
(2) the Indo-European Proper neo-root (attested embryonically for this root!) offers more
explicatory puissance than the tenuous connections with the Gothic verb of a different
semantic sphere, usually assumed. Solta (1960: 105) assumes the same ablaut for the
Armenian cognate: tam (< *da-ie-mi), etu (< *e-do-m), respectively ‘I give’, ‘I gave’. The
alignments of the aorist/preterite here is not without interest: Arm. etu, Skt. d-dam
(*e-dom): Gk. éôuxa, ON tok (*[e]-do-K-), no corresponding alignment existing in the
present.”

3. The Fata Morgana of Anatolian + (-K)

A perusal of the entrenched allegations cannot be eschewed, but must be executed in
fashion most summary and limited to those germane to the exhibits of this paper: — (1) in
a drawn-out attack on ‘the etymological bases’ that Sommer used to understand tak3ul and
takSan, to wit derivation fromv¥tekp-, Pedersen (1938: 139—44) canonized a prima facie im-
probable derivation of takSul (and words which he allowed to be connected) from neo-
Ydbek- +, while he consigned tak3an to Vdwo- +: “Allem Anschein nach ist /dakkeszi/ mit

lat. facesso identisch". Pedersen was overly taken with the Phrygian comparison a66axer/
eSaec, the aorist being a reasonable ‘structural’ parallel to Hitt. dai. He analyzed tak3-as
having the -k seen in Latin and Phrygian and the Hittite stem as a whole as having this same
-k + the -s ‘Erweiterung’, the ‘combination’ seen in Ven. vhaxy:s: vo. Unfortunately, his
semantic comparisons are thin; the morphological segmentation is horrible; the phonology
leaves much to be desired; and Phrygian is only geographically — not genetically — Anato-
lian.
Pokorny has the best of all worlds in dividing the meanings under vdhe-: *Dieselbe k-Erw.
. auch in... (dakkeszi) ‘macht, stellt hin’ . . . (daksul) ‘freundlich’” and Viekb-: “takš-,
takke¥- sammen igei, unternehmen’.” fran (1917: 56, fn.2) had suggested that takkul
could perhaps belong to the root of Skt. dáksati, Lat. decet (thus IE Ydek-). For takÿan
Hrozny (1919: 63, fn.9) had suggested “in Reihe", an -so stem from the root seen in Gk.
rat (IE *fg-, in which he conjoins what Pokorny gives as two roots, fag- and fag-).
Watkins (1969: 117) construes tak¥- as being an extension within Hittite of takki, which
had been etymologized by Laroche as < *dok-*4: ; this latter is an o-grade of the root seen
in Gk. Sek7ro and the extended form (< *doks-)i is reminiscent of Gk. ŝota (thus, IE Ydek- ).
Scholarly decency would have been well served if the Doctor Cambrigiensis had deigned to
take note of the proposal of the Pioneer. (2) It was Sturtevant who canonized fekkuÿ- as
a Hittite cognate to IE neo-vdeik-.?5 In his article (Language 6/3. 1930: 228) he refined his
analysis (of Language 6/1. 1930: 27) with the following formulation: “Hittite tekkux- .
represents IH **deikus-, i.e. IE *deik-plus the element u/u ... plus the formative s, which
gave rise to the IE present and aorist suffixes in s” (thus neo- -Vdeik- + +, and in tekkuS($}-
anu-, we can add another ‘+’). Feist (s. ga-teihan), in citing Sturtevant’s analysis, includes
an apparently personal communication from E. Forrer: “bezweifelt von E.F.: die Bild.
scheint luw.”. 44 years later one is still tantalized by Forrer’s grounds for perceiving a
Luwian structure.

In 1974 Puhvel (reprinted 1981b: 263) uses this word as an exhibit for the Hittite reflex
of medial preconsonantal *kW~: “tekku38ai-/dekWsai-/, tekku¥(Sa/nu- /dekWsnu-/ . .
cognate with Avest. daxš- ‘teach’, daxSta- ‘sign, characteristic’. The Avestan words are
otherwise isolated (unrelated to Skt. déks-..., dasasydti ..., Lat. decus, IE *dek-) but
point clearly to a labio-velar ..., whereas the standard comparison of tekkus($)- with IE
*deyk- ... must operate with an unexplained suffixal *-us-.” Watkins (1969: 229) speaks
of a thematic -s suffix to a stem in -u, thus *tek-u-se/o-, from a root *tek- (seen in Gk.
rékuap[rékucop) as there is neither phonological nor morphological justification for the
traditional comparison with the Greek and Latin -5 aorists, éóet£a, dixi.

In utilizing the tradition of dakke$- < *dlok- +, and tekkuSai- < *deik- ++, to buttress
her own argumentation of a “coalescence de deux élargissements" in the IE suffix -ske/o-,
i.e. an agglutination of independent morphemes, allegedly seen in reverse order in da-k-
(e)*-, etc., my dear friend, F. Bader (1974: 35), is relying — quite irrespective of the merits
of the argument as a whole — on the support of the proverbial fragile reed (here not Egypt,
but putatively Hatti). (3) Hoffner (1977: 105, 107) determined the meaning of Hittite
ekt- as '(hunting) net' and, with Berman, concluded that this item was the Nesite cognate
of aggati-, “the Luwian word... based upon the same Indo-European root (perhaps the
verb *ye- with -k- enlargement; cf. Latin iacio, ieci ...)...”. Hamp (1978: 120) laconical-
ly noted: “This well known root ... would not be expected (outside of Latin, where we
also find facio feci) to show the generalized -k-, particularly in such a nominalization . . .

therefore we do better to seek a root terminating in a velar (or Indo-European palatal) (my
emphasis). ... I propose that we now have a new IE root *iek-, or *ek”- ‘hunt’”. In 1972
(reprinted 1981b: 222) Puhvel had likewise (per the ex-meaning of the word, ‘leg’) pro-
posed a derivation from the neo-Vey-gh- (proto-root *ei-) 'go'.?9 In a neo-footnote (1981:
414, fn.41) he now offers ekt- ‘net’ as cognate with Lat. ictus, < *ayk-t-. This is Pokorny’s
aik- : ik- and seems to be the same root as Hamp's "jek -, thus *ayk- -/yek-/ik- (I would say
*ayek-, but fear to, lest any man segment it as ixi. ‘give’ + -k, with a semantic develop-
ment ‘give the spear/net to’!).
A chart will summarize the proposals:

tak(ke ekti-Jaggati
Proposed | Alleged Neo-Roots: - **vek-, **eigh-

IE Reasonable proposed v’s kb- *tek-

Real Roots (IH Proto-v’s) ç- *dekW-

My object in this section is not to weigh the recent reasonable proposals, but only (1) to
emphasize that, whatever roots may be involved in the respective Anatolian words, they are
certainly not the IEP neo-roots *dhak-, *deik-, and *yek-, and (2) to demonstrate the
debacle effectuated for Anatolian etymology by the non-recognition of the -[-K] factor
in Anatolian vs. its extancy and productivity in IEP, in whose early stages its Wuchern (the
connotations of which German “articular infinitive” English cannot express) had its
genesis.??

Notes

The subtitle is, of course, a take-off on “Multiple Stem Conjugation: An Indo-Hittite Isogloss?” by
my lamented friends, J. Alexander Kerns and Benjamin Schwartz (1946), zixrönam livräxä. In cita-
tions, every attempt has been made to reproduce the style of the quoted author, e.g. titalics, tdia-
critics, etc.

For the terminology and the thereby implied perspective, see Adrados (JJES 10/1—2: 1982).

See my articles 1974: 70-76 and 1976: 79-80. I am preparing one further article on this verb
(with, inter alia, reference to its extra-IH connections), which is already cited in Bomhard (1984:
203).

A speculative, but undemonstrable, perspective within this choice could maintain that Anatolian
alone preserves so great an archaism of ur-vocabulary that all traces of the item have vanished from
all cousin languages. Such reasoning, though possible, leads nowhere.

Those of my school would conceptualize this choice as Anatolian having preserved a more primitive
(1) configuration (formally) and/or (2) meaning (functionally). This conceptualization could (for
him who so wills) be oppositely set, to me an unlikely analysis of relationships.

© On pe- sec my view 1974: 71 where one of these reasons is already mentioned. I am grateful to G.J.
Ayala for having pointed out to me that that article would have been more logically entitled ‘Why
‘Two Adverbs ... Became Inseparable (Hence, Preverbs).”

I have not seen the article cited by Weitenberg, but Rosenkranz repeats the view in his book (1978:
37 38), where “pai- ‘geben’ zu ep/app- ‘nehmen’”’ is considered under the heading of quantitative
ablaut. We are told that, in the listing of these examples, “‘seien einige bisher kaum beachtete Bei-

18
19

spiele für ein Nebeneinander von Status I und II der Wurzel vorweggenommen, sämtliche Ableitun-
gen mittels des Suffixes -ai-/-ija-". Besides the semantic tenuousness of most of the examples, I find
the phonological principles beyond my grasp: if the Hittite ablaut epzi/appanzi < *aep-/gp- (the
second Hittite form < ¢-grade, with vocalization of the laryngeal vel sim.), can the same $-grade
*opai- 2 pài-? This latter process seems to be that which occurs e.g. in Skt.: dsti/sdnti < *gésti/
2sönti, but precisely that which does not occur in Hittite, where the same reconstructions > ekzi/
aganzi. Perhaps Rosenkranz has in mind Gamkrelidze’s (1968: 89) metathetic laryngeals: *egp- >

ep-, #5 >app-, *apai- > pai-, to be sure a variation even on Gamkrelidze’s view. Re Cop and other
views, see fn. 11.
The rest of the sad tale of the proposals for iwaru is well summarized in Tischler (1983:455).
The first alleged isogloss does not concern this paper.
Petersen's preverb-root division seems to leave no root, only preverb (pe-) * desinence (-hhi-). The
history of the proposals re dai- ‘place’ and da- ‘take’ is — if possible — more obfuscationistic than the
pai- proposals.
To properly understand the context in which *p-Yai- came to be crowned, a historiography would be
necessary. For such (and equally for the matters of fns. 7 and 10), there can be no proper place
here. Most of the “doubters” base themselves on the grounds mentioned in fn. 13.
Personal correspondence, dated 26 April 1977. Credit goes to Prof. Cowgill for the framework, but
he — of course — bears no responsibility (for worse or for better!) for that which I will here attempt
to build on his foundation.
In juxtaposing the same set, Watkins (1962: 100) gives the following chart (relevant for his purposes
there, sadly disappointing for understanding Anatolian pai-/piya-):

*dheH-s (Skt. dhas) : *dheH- (Skt. dha-)

*Hei-s (Hitt. p-ai¥ ‘gave’) : *Hei- (Hitt. p-ai).
The grounds for dubiety of almost all the doubters are a rehashing of the old debate between Sturte-
vant and Couvreur concerning Hitt. a = IE a being a reflex of the ad hoc established 4th laryngeal.
CowgilPs objections, thus, accord with (1) this universally repeated caveat and (2) the conjugational
parallelism of pai- with dai-, as noted e.g. by Puhvel (1960: 55--56, where he is as ‘non-committal’ in
the opportunity presented for a self-extrication from *p-YGi- as he charges Risch with being re the
dái- type as a whole) and Kurytowicz (1979: 146).
In some individual words and categories we may have Verschärfung or other kinds of reflexes of
former laryngeals; but the PIH I envision was a whole lot more that the sum of laryngealistic grunts.
This PIH (> PIE and Pre-Anatolian) had a morphology, had suffixes, and —as such -- was a language,
not an assemblage of primaeval hunting calls.
The summary and views of Poultney (1959: 131—32) on the pres. subj. 3sg. feia and related forms
are critical for our analysis: These forms have been referred to the full grade of the root *dhe- . . . ,
sometimes to the extended *dhe-k . .. In favor of the second analysis are . . . and the fact that this
root without the -k extension is not positively known to have existed in the Italic languages [in the
simplex] ". References are provided to the controversy between Von Planta (X. *fe-) and Buck
(< *fé-k-). Poultney's second support for the < *dhe-k analysis is, alas, a less than reliable testi-
mony, if we conceptualize the process as productive, alive, and well throughout IEP and if we add to
this the factors discussed in footnotes 16--19 below.
Beeler (1981: 67): “Another [significant lexical isogloss of Venetic with **Italic"] is the occurrence
of the verb *fak-”, citing these Latin, Oscan, and Umbrian forms.
Polomé (1966: 59-76) declares against a close Venetic-Latin genetic relationship, much of his article
being devoted to old views of Beeler (adumbrated, on p. 52 in an article on other matters in that
same volume, by Beeler, pp. 51-58). In his most recent declaration on the relationship or lack
thereof, Beeler (1981: 65—71) comes, after 40 years of struggling with the data, *to the disheartening
conclusion that the problem is essentially insoluble. . ."
See Beeler (1966: 51—58) re a languages-in-contact explanation in preference to a Proto-Italic.
The -do palaeo-compounds (cf. fn. 15 above) are not relevant synchronically. But beyond this — and
more important — it is questionable whether, without the IE comparative materials (esp. the Gk.
t [-«] stems), even the diachronist would have connected these compounds with facio rather than with
do 'give'. There is an important distinction here in the possibilities for the diachronist as opposed to
the comparativist, the former being sensu stricto scientifically monoglottal. On the native under-
standing of the palaeo-compounds (< ah) -), sce Ernout-Meillet (1959: 179--80). I find Kronas-

ser’s (1956: 166; 197) comparisons of these Latin compounds with putative Hittite compounds,
allegedly built to the Hittite cognate of the Latin root, unacceptable (see my article on “ie. RHA
31. 1973: 104). The observation of Ernout-Meillet (1959: 304) that, as opposed to the situation
whereby we have palaeo-compounds in -do as well as neo-compounds in -ficio, in the case of iacio,
all compounds are of the con-ficio type, none of the con-do type, is indicative once again of the
on-going-ness of the process. The causative factor for the absence of the *con-io type ll con-do type
is obviously that all one would hear/see is preverb (con-) + desinence (-io, as in e.g. capio, audio); in
the important 1sg., there would be no meaning-bearing root.

I cannot here discuss any further sense developments. Suffice it to say that I tend to concur with
the further connections considered by Pokorny: 507, 508 & 510, only with a different conceptuali-
zation of the interrelationships.

The survival of *yewos in Lat. ius is a factor of the archaic juridical nature of the item. Concerning
the meaning ‘the fit/meet’ | Skt. rta-, Hitt. ara et al., see my comments (1981: 906—12 et passim)
and (1982: 34—35). There is a similar semantic archaism in German ‘mit Fug und Recht’.

The old meaning of Norse selia, viz. ‘give’, survives in Mod. Eng. handsel ‘a present to express good
wishes’, originally ‘a giving of the hand’. The former meaning of Mod. Eng. sell is still seen in the
Anglo-Saxon Gospels, e.g. Mark vi:25: uolo ut ...des mihi... caput Iohannis rendered Ic wylle
pet du me ...sylle iohannes heafod (other mss. selle [see the ms. and dialect distinctions in Skeat
(1871)1). The competing Germanic lexical item occurs, inter alia, in the Gothic of the same passage:
Gk. @éAw iva … 56 wot ... TRY kepaññv 'Iwävvou rendered wiljau ei mis gibais . . . haubib Iohannis.
While concurring for the 3sg. (et = Skt. d-dat < *e-dot), Godel (1975: 126) insists that the preserva-
tion of the presence of a root vowel in the other persons (e.g. Solta’s eru) demands the at one time
extancy of a subsequent syllable. He excludes any comparison of this lsg. with either Skt. ddam or
with Gk. €5wxa and favors a derivation of etu from *[e-] dosom (comparing -s aorist forms of OCS:
daxii and of Alb.: daSé). Meillet (1936: 133) offers an anomalous preservation < *etuy. Kerns and
Schwartz (1972: 61) offer an overly summary aside: “...a few whole paradigms as sg. 1 edi, etu
seem recast from IE athem. root aorists *e-dhe-m, *e-do-m...” The niceties of the Armenian con-
jugation as a whole need not concern us here, provided we limit — for present purposes — the com-
parison to 3sg et with Skt. d-dat (IE/IH possibly *[e-]d0-t/s (Godel 1975: 126).

Note the remarks of and references in Melchert (1977: 102 and 105 with fn.12) concerning (1)
Laroche’s etymology; (2) connection of Anatolian palaeo-Vdak- with Toch. A. task-, B. tas- (< *tak-
sk-) ‘resemble’, and (3) Van Windekens' 1941 (!) connecting of the Common Tocharian item with IE
vdek-.

The present article had been conceived as having two halves, the second of which concerned the
Hittite suppletive verb /de-/, /tar-/ ‘say, speak”. In the event, the semantic typological data became a
tumulus which, D.v., I will present elsewhere. In the meantime, Puhvel (1982: 183, fn.11) has
written that the view that Hitt. /de-/ continues the IH proto-root, seen with + [-K] extension in IEP
neo-vVdeik-/$-, “found its latest avatar in Y. Arbeitman”, with reference to my article in RHA 31.
1973: 104ff.. We would have in this poetry a literal case of Er verbum caro factum est if only
verbum meant ‘verb’ rather than ‘word’. Suffice it to say here that the correct cognates of Hitt.
/tar-/ are given in Pokorny 1959: 1089 (spec. the Balto-Slavic items listed), whilst I now see the IEP
cognates of Hitt. /de-/ as consonant with the scheme which has been sketched in this paper. As such,
it would seem to require no avatar or other extra-natural means. At all events, flattered as 1 am by
this logo-theology, my frail vessel is unworthy to serve as the transient dwelling of an IH root. It is
not lacking in germaneness to note that Puhvel, /ui-méme, has apparently abandoned his adherence
to tekkuS(a)nu- et al. X (m.s. IEP neo-V)*deik- (his position in 1960: 26). None of us is proposing any
connection between Hitt. /de-/ and rekku¥..

The etymology for the meaning ‘leg’ seems to originate with Kronasser. See the summary in Tischler
(1983: 351—52; 349—50!).

It is startling that Georgiev and I have in common but one variable in the respective etymologies we
have proposed for Phryg. Bexos ‘bread’, to wit an IEP neo-Vformed with a + [-g]. I interpreted it
(1981: 973) as that which The Distributor/Giver (sc. God) distributes/gives us to take/eat (payei),
“Distributor, give, eat” all being cognates of Bekoc from neo-vblYe)EY-äg-. Georgiev (1981: 131)
analyzes it as cognate to Eng. bake, Gk. yuoyw, etc. This latter is, likewise, an IEP neo-V, built with
a * [3] to the proto-root phe. “to warm’, scen in Germanic "beóam ‘bath.’ Puhvel (1979: 213,
reprinted in 1981b: 360) notes a Hitt. noun /bita/ (my phonemization) ‘gift, grant, allotment’, in one

locus: ‘he eats [his] bread-allotment (NINDA-LAM pitta)’. Puhvel analyzes this noun as < *piyatta,
nom./acc.pl.nt. From the formal perspective, there is a neat proportion between the verbs proposed
respectively by Georgiev and me and their respective nouns:

IEP Neo-V Proto-Root Noun i
*bligg- (< *blEY.dg-) *blitóm (< *blivotom < *bl'yotom « *bNEYotom) “Gift/Por-
*ble/og- (« *bheg -g-) *bDatém “Warmth/bath”. tion”

References

Arbeitman, Yoél L., 1974, “Why two preverbs (and only these two) became inseparable in Hittite”,
JIES 2: 70-76.

Arbeitman, Yoél L., 1976. *Anatolian Piya-'give’”, JIES 4: 79-80.

Arbeitman, Yoél L., 1981. “The Hittite is thy mother: An Anatolian approach to Genesis 23”, in Bono
Homini Donum, edited by Y. Arbeitman and A. Bomhard (Amsterdam: John Benjamins B.V.),
889-1026.

Arbeitmann, Yoél L., 1982. “Luwio-Semitic and Hurrio/Mitannio-Semitic Mischname-Theophores in the
Bible, on Crete, and at Troy”, Scripta Mediterranea (Bull. of the Soc. for Med. Studies, Toronto)
3: 5—53.

Bader, Françoise, 1974. “Persée, mépôw et l'expression archaique du temps en indo-européen”, BSL
69: 1—53.

Beeler, Madison S., 1966. “The interrelationships within Italic”, in: Ancient Indo-European Dialects,
edited by H. Birnbaum and J. Puhvel (Berkeley: University of California Press), 51-58.

Beeler, Madison S., 1981. “Venetic revisited”, in: Bono homini donum, edited by Y. Arbeitman and A.
Bomhard (Amsterdam: John Benjamins B.V.), 65-71.

Bomhard, A., 1984. Toward Proto-Nostratic (Amsterdam: John Benjamins B.V.)

Carruba, Onofrio, 1966. Das Beschwôrungsritual für die Gôttin Wiurijanza (= Studien zu den Boÿazkôy-
Texten 2) (Wiesbaden: Harrassowitz).

Ernout, Alfred/Antoine Meillet, 1959. Dictionnaire étymologique de la langue Latine, 4th edition
(Paris: Klincksieck).

Feist, Sigmund, 1939. Vergleichendes Wörterbuch der gotischen Sprache (Leiden: Brill).

Gamkrelidze, T.V., 1968. “Hittite and the laryngeal theory”, in: Pratidanam, edited by J. Heesterman et
alterae (the Hague: Mouton), 89-97.

Georgiev, Vladimir I., 1981. Introduction to the History of Indo-European Languages, 2nd edition
(Sofia: Bulgarian Academy of Sciences).

Godel, Robert, 1975. An Introduction to the Study of Classical Armenian (Wiesbaden: Reichert).

Götze, A./Holger Pedersen, 1934. Muršilis Sprachlihmung (Kébenhavn: Levin and Munksgaard).

Hamp, Eric P., 1978. “Hittite ekt- (hunting) net ^, JF 83: 119—120.

Hoffner, Harry, A., 1977. “Hittite Lexographic Studies, 1”, in: Essays on the Ancient Near East, edited
by M. de Jeffrey Ellis (Hamden: Archon Books).

Hrozny, Bedrich, 1917. Die Sprache der Hethiter (Leipzig: Hinrichs).

Hrozny, Bedrich, 1919. Hethitische Keilschrifttexte aus Boghazkôi (Leipzig: Hinrichs).

Kerns, J. Alexander/Benjamin Schwartz, 1946. “Multiple Stem Conjugation: An Indo-Hittite Isogloss?”,
Language 22: 57-67.

Kerns, J. Alexander/Benjamin Schwarz, 1972. A sketch of the Indo-European finite verb (Leiden: Brill).

Kronasser, Heinz, 1956. Vergleichende Laut- und Formenlehre des Hethitischen (Heidelberg: Winter).

Kurytowicz, Jerzy, 1979. “Die hethitische hi-Konjugation”, in: Hethitisch und Indogermanisch, edited
by E. Neu and W. Meid. (Innsbruck: Institut fiir Sprachwissenschaft der Universitit Innsbruck),
143-146.

Meillet, Antoine, 1936. Esquisse d'une grammaire comparée de l'arménien classique. 2nd ed. Vienne:
Mekhitharistes.

Melchert, H. Craig, 1977. “Tocharian verb stems in -tk-", KZ 91: 93—130.

Pedersen, Holger, 1938. Hithitisch und die anderen indoeuropäischen Sprachen (Köbenhavn: Levin and
Munksgaard).

Petersen, Walter, 1933. “Hittite and Tocharian”, Language 9: 12-34.

Pisani, Vittore, 1974. Crestomazia indoeuropea (Torino: Rosenberg and Sellier).

Polomé, Edgar, 1966. “The position of Illyrian and Venetic”, in: Ancient Indo-European Dialectics,
edited by Birnbaum, H. and J. Puhvel (Berkeley: University of California Press), 59-76.

Poultney, James Wilson, 1959. The Bronze Tablets of Iguvium (Baltimore: Johns Hopkins University).

Puhvel, Jaan, 1960. Laryngeals and the Indo-European Verb (Berkeley: University of California Press).

Puhvel, Jaan, 1981a. “The meaning and source of Hittite fuwaru”, JAOS 101: 213—214.

Puhvel, Jaan, 1981b. Analecta Indoeuropea (Innsbruck: Institut für Sprachwissenschaft der Universität
Innsbruck).

Puhvel, Jaan, 1982. “Baltic-Anatolian Lexical Isoglosses", in: /nvestigations philologicae et compar-
ativae, edited by E. Neu (Wiesbaden: Harrassowitz).

Rosenkranz, Bernhard, 1978. Vergleichende Untersuchungen der altanatolischen Sprachen (the Hague:
Mouton).

Skeat, W.W., 1871. The Gospel according to Saint Mark in Anglo-Saxon and Northumbrian Versions
(Cambridge: Cambridge University Press).

Solta, Georg Renatus, 1960. Die Stellung des Armenischen im Kreise der indogermanischen Sprachen
(Wien, Mechitaristen).

Tischler, Johann, 1983. Hethitisches etymologisches Glossar, part 1 (a-k) (Innsbruck: Institut für Sprach-
wissenschaft der Universität Innsbruck).

Watkins, Calvert, 1962. Indo-European Origins of the Celtic Verb 1 (Dublin: Dublin Institute for Ad-
vanced Studies).

Watkins, Calvert, 1969. “Formenlehre, Teil 1”, in; Indogermanische Grammatik 3, edited by Jerzy
Kurytowicz (Heidelberg: Winter).

Tranica in Caucasian

Harold W. Bailey (Queens’ College, Cambridge)

The very large complement of early Iranian loan-words in the Kartvel Georgian language
has been collected up to the 12th century by Mzia Andronikasvili (1966). Similarly A.
Genko (1930) examined the Iranian words in the Ingu$ vocabulary. Scattered among the
Caucasian languages are other Iranian words which are important for the still scanty early
Iranian vocabulary. Some of these correspondent words had been put together in Bailey
(1975: 31—35). A similar statement on è gni has been published in Georgian (Bailey 1982:
99—101). Here are six words which are a distinct contribution to Iranica.

1. Kartvel (Georgian) c ’ign-i, c’ignak”i ‘book’ has been often canvassed!. But the connected
early Iranian has been overlooked.

Iranian has Zoroastian Pahlavi, New Persian &ak ‘document’, written both #k” and &y’, that
is, tak and tay, in Zoroastrian Pahlavi. This word is in the Sasanian Legal Code the Matiyan
i hazar datastan ‘Book of the thousand decisions’? . This čak was met by C. Bartholomae in
his study of the legal text?, and it is now in the two editions of the text by A. Perikhanian
(1973) and M. Macuch (1981).

The Iranian Zak can be directly from a base &ak-, as tak- ‘swift, bold’ is directly from tak-
‘to run’ and tak- ‘be bold’. But the ZorPahl. &y *tay may show rather a base tag-, whence
in tak the final of *taga- has been replaced by final -k.

The participle to tag- is tagna-, later tigna-, whence *tayna-, *tiyna-. It is attested in a
compound with preverb pati-, pati- in Manichean Sogdian p’téynyy “patitagnaka- ‘copy’.
As a loan-word this had long been known in Hebrew ptXgn, and with pr- for pt- in Aramaic
pršgn, Syriac parSagn-a rendering LXX 'avriypaġov ‘copy’. The corresponding Parthian
gave Armenian patten, and the word is in Zor.Pahl. paten (or patten ‘copy’), ham-paten,
and Pazand patin (SGV 16.24).

This tagna- is the form of the participle, as bagna- is from bag- ‘to bestow’, Sogd. Byn- in
prByn ‘gift’. It has passed to bigna- in Old Persian bigna- of the names Bagabigna- and
Ariabignes.

This Yagna- without pati-, pati-, is preserved in Dido Segen ‘book’, Veinakh Batsbi jagn,
Zagn, Ceten jaina, Zaina, plural Zaina; Ingu¥ žai, plural Zaina’.

The replacement of the tagna by tigna- shown the early change of ta- to ti- as in Avestan

tanah-, Cinah- Soy’, and in Old Pers. #inah- in the name ‘Aomaëwnc. Kartvel Georgian has
this £i- in the word c ign-i ‘book’ (with ci- from older 2).

2. Kartvel Georgian nask v-i ‘knot’, verbal nask va, has the -v-i common in loan-words, such
as, nadzy-i ‘fir tree’, New Pers. naz, Ossetic nazu, Iron naz ‘fir’ and nézi, Iron nézy ‘pine’;
mogv-i ‘shoe’, Zoroastrian Pahl. mok, Abkhaz a-magu; nest’v-i ‘trumpet’, verbal nest ’va,
nest i; nesy-i ‘melon’, Ossetic nasd, Iron nas ‘pumpkin’, Ingu¥ nars ‘cucumber’ (to Ossetic
nars- ‘to swell’).

The Avestan religious technical term naska- of the twenty-one sections of the original
Avesta was taken over into Zoroastrian Pahl. nask, the southern dialect says na¥k (as in
Avestan araska-, NPers. araSk ‘envy’). This Kartvel nask* can thus support ghe connexion
with the base nad- ‘to bind’ as in Bartholomae (1904).

The name of the Aragvi river is already in Strabon, as Aragos (gen. 'Apayov) and Aragon
(acc. 'Aparyciva..

3. Čečen varš, plural vara, and gen. var¥an (used also as an adjective) is rendered ‘grove,
thickets’ in the phrase varkan jistie jolu (j = i) ‘at the edge of a wooded slope’ (Jakovlev
1940: 304).

With this var¥ can be set Avestan varaSa- (glossed by Zoroastrian Pahl. vesak ‘a wood’)
‘some kind of plant’ and vara%gji- glossed by Zoroastrian Pahl. ré¥ek ‘root’, bun ‘stalk’ and
dwn, Pazand eyvan ‘upper part of a tree’, Parsi Sanskrit skandha-. In modern Iranian the
archaic Yazgulami has warX ‘a herb’, and Suyni wart ‘a mountain grass’, from *varta-.
From New Persian the AIW quoted varSan ‘wood pigeon’, Syriac warsan-a, Akkadian
amurSanu, urSanu (amu-, u- for foreign u).

In spite of its ambiguous -ks- (which corresponds to four Iranian sounds xš, 5, yz and 2)
the Old Indian vrksá- can be brought in here. The meaning of later Indian is ‘tree’, but for
the Veda the word is also used of the dndhas- ‘herb’ of the soma plant (the original meaning
of andhas is clear in RV 1.28.7 of the horse’s fodder). In RV 10.94.3 vrksdsya sakham
arundsya is ‘the sakha (stalk) of the reddish herb’ rather than ‘tree’. A parallel for an earlier
Vedic meaning can be seen in RV 1.182.7 mrgasya pataror ‘of the flying beast’, that is
‘bird’, where later the mrga- is an animal as RV mrgó hasti ‘elephant’ or ‘deer’. In Iranian
the mrga- is ‘bird’ throughout except Wakhi merg, marg female ‘wild goat’.

Ossetic bdlasd, Iron bálas ‘tree’ is to be taken from *ulaSa- with the vowel shift as in Khota-
nese Saka grama- ‘hot’ (beside first component garma-), Middle Parthian (Turfan) grb from
garba- ‘womb, embryo’, Middle Persian (Turfan) hrg *hray ‘tax’ with Khotanese Saka
harga- ‘tax’, Armenian loan-word hark, Kartvel Georgian xark -i, Arab-Pers. xardy.

It is possible to bring these various words together to the Proto-Indo-European base uel-
(Pokorny 1139) under welk-, Avestan varasa- ‘hair’, Zoroastrian Pahl. vars, Armenian loan-
word vars, New Pers. gurs, to Old Slav. viasii; uelk- Av. varaka- ‘leaf’, Zoroastrian Pahl.
vark, valg, New Pers. barg, Khotanese Saka baggara-, Sogdian, Parthian wrkr; Old Ind.
valkd- ‘bark’, vdlsa- ‘twig, sprout’. To this the enlargement -s- gave Iran. varSa- and vrSa- and
Old Ind. vrksá-.

4. Avar parss and parssi* , plural parssal, parssabi ‘rock’ is of valuable support to the Khota-
nese Saka pase ‘pieces of rock, stones’ in Jätaka-stava 13 v 2 pase te rriscye gihaifid
hvastamdd ‘they struck sharp cutting stones upon you’. This was missed earlier in my
Dictionary? . The pase 'stones' will be plural to pasa- from an a earlier *pasa- with subscript

hook, from *palsà- to Proto-Indo-European pel-, from pel-k- (like Khotanese-Saka pasa-,
older pgsa- ‘messenger’, base pars- ‘to send out’) beside pel-s- in PaSto parSa, Kati (Nuristani)
pari, Old Ind. pas- in pasyá-, pasana-, Old High Germ. felis, Old Norse fjall.

5. The Iranian base tap- ‘to strike’ is in Zoroastrian Pahl. rapah ‘injury’, tapahenitan ‘to
destroy’ (glossed by kahet ‘to reduce’), New Pers. tabah, and in Iranian tapar ‘axe’, Zoro-
astrian Pahl. tapar, Armen. loan-word tapar, Middle Persian (Turfan) tbr, tapar, New Pers.
tabar, Čeremis loan-word taßar, Slavic: Russ. topor ‘axe’. Old Slavonic has tep- ‘to strike’.
This tapar is formed by suffix -ar-a-. A confirmation of this connexion can be seen in the
Tabarsarani fabus, tawux (Y and Y4) hammer (the inflected forms tabšu- and tovš- are cited
in Erckert (1895: 912) 3% as rabX4- which reflects the suffix -uXa- used of tool and agent.
Avestan has taku¥a- (gen.plural takuXangm) beside taku- in the plural &akavo 'hammer, axe'
as a weapon, New Pers. faku¥. It occurs besides forms with the suffixes -ufa- and -uka-, as
in Zoroastrian Pahl. čkwč *ťakuč, Turk loanwords čäküč and tdkiik. The base is čak- as in
Ossetic cayd ‘struck’, caxtà ‘he struck’, present -d- cäydun. The Guzz Turk čäkük is in Al-
KaSyari. Kartvel Georgian has tak ‘ut‘i and Ingus$ Caqut. This same -u3a- suffix is found in
Old Indian in purusa- (the adhibhavayitar ‘secondary nourisher’ following bhavayitri the
mother), connected with the base par- ‘to nourish’.

The Tabarsaräni has also from the base bar-, bur- ‘to cut’ a word *buruSa- ‘shears’ preserved
as ubruxH.

The Caucasian tabu¥ confirms the derivation of tapar from fap- ‘to strike’ and excludes a
metathetic connexion with *para0u-, in Ossetic fürdt ‘axe’, Khotanese Saka pade ‘axes’
from *paratu-.

Notes

G. Deeters (1965: 53--56) did not connect cigni with Iranian, and it was after all not a wandering
word without basis. He rightly set aside any connexion with Latin signum.

The word matiyan is the Sasanian form of older m'tgd'n, m'dgd'n *matakadàna- in Parthian form,
from which the -akada- has been developed to -iyan-, as *matakadara- ‘steward’ has survived in Ar-
menian matakarar, Zor. Pahl. måtiyār (m tT).

From matiyan Armenian took matean, gen. mateni, and thence came Kartvel (Georgian) mat 'iane.
The same -iyär came from -adára- in New Pers. Yhriyär.

3 Bartholomae (1904: part 1: 8, 21; part 3: 30; part 4: 10, 13, 17).

^ Zirkov (1936) has also parass, but it is not to be found in Saidov (1967).

5 For amendment to the Dictionary of Khotan Saka see Bailey (1981: 15).

References

Andronik'a$vili, Mzia K., 1966. [Studies in Iranian — Georgian Linguistic Contacts] 1 (Tbilisi: Tbil.
Univ. gamomcemloba).

Bailey, Harold Walter, 1975. “Excursus Irano-caucasicus”, in: Acta Iranica 4, dedicated to H.S. Nyberg
(Teheran-Liège-Leiden), 31-35.

Bailey, Harold Walter, 1979. Dictionary of Khotan Saka (London: Cambridge University Press).

Bailey, Harold Walter, 1981. “Indo-Iranica”, Indologica taurinensia 9 (in memorial to L. Sternbach)
(Torino: 1981), 15-18.

Bailey, Harold Walter, 1982. “Iranuli CAK, CAGNA 'sabut'i! — k'art'uli ‘cigni’ (Iranian Cak, Cagna
‘document’ — georgian c’igni ‘book’)”, in: Macne (Tbilisi: Tbil. univ. gamomcemloba), 99-101.

Bartholomae, Christian, 1904. Altiranisches Worterbuch (Strafsburg: Trübner).

Bartholomae, Christian, 1918. Zum sasenidischen Recht 1 (= SbAdW HD 9/5,14) (Heidelberg: Winter).

Bartholomae, Christian, 1920. Zum sasanidischen Recht 2 (= ShAdW HD 11/18) (Heidelberg: Winter).

Bartholomae, Christian, 1922. Zum sasanidischen Recht 3 (= SbAdW HD 13/5) (Heidelberg: Winter).

Bartholomae, Christian, 1923. Zum sasanidischen Recht 4 (= SbAdW HD 14/9) (Heidelberg: Winter).

Deeters, Gerhard, 1965. "Ein weiterer Beleg des osteuropäischen Wanderwortes für “Buch””, IF 70:
53—56.

Von Erckert, Roderich, 1895. Die Sprachen des kaukasischen Stammes (Wien: Hólder).

Genko, A., 1930. Iz kul’turnoj istorii inguSev “[From the cultural history of the Ingushes]’, Zapiski
kollegii vostokovedov 5 (Leningrad:

Jakovlev, N.F., 1940. Sintaksis fe&enskogo literaturnogo jazyka [Syntax of literary Chechen] (Moskva/
Leningrad: Akademija Nauk SSSR).

Macuch, Maria, 1981. Das sasanidische Rechtsbuch 2 (Wiesbaden: Steiner).

Perikhanian, Anait Georgievna, 1973. Sasanidskij sudebnik [The Sasanian lawbook] (Erevan: Akade-
mija Nauk Armjanskoj SSR).

Pokorny, Julius, 1959. Indogermanisches etymologisches Wörterbuch 1 (Bern/München: Francke).

Saidov, Magomed Said, 1967. Avarsko-russkij slovar’ (Avar-Russian dictionary) (Moskva: Sovetskaja

t enciklopedija).

Zirkov, Lev Ivanovis, 1936. Avarsko-russkij slovar’ [Avar-Russian dictionary] (Moskva: Sovetskaja
enciklopedija).

Prefixal negation of English adjectives: Psycholinguistic dimensions
of productivity

Philip Baldi (The Pennsylvania State University)

1. Introduction

It is the purpose of this paper to provide some preliminary results of ongoing psycholin-
guistic experimentation designed to test the productivity of certain negative prefixes in
English adjectives. Despite their preliminary nature, it is hoped that the exposure of these
data will stimulate comments and criticisms from colleagues in both psycholinguistics and
historical linguistics which will help us to refine both the data and the methodology of the
inquiry.!

To undertake a psycholinguistic study as a means of approaching morphological produc-
tivity is, in my opinion, a fitting tribute to one of the most important figures in the history
of American linguistics, Henry Hoenigswald. Hoenigswald is a special blend of scientist and
philologist which is fast becoming a rarity in our profession, and it is an honor to present
this paper to him in recognition of his place in the field of linguistics.

The issues of this study are familiar to historical linguists and to those whose interests lie in
the linguistic history of English. Specifically, I am concerned with the productivity of the
negative prefixes un-, in- (im-, il-, ir-), as well as non- and, to a lesser extent, dis-. As we will
see below, un-, in-, non- and dis- are the most productive negativizing prefixes in English
today.” We are investigating not only the productivity of the specific prefixes, but also
how the patterns of productivity are determined. That is, how do speakers choose a specific
negative prefix for a specific root?

In approaching the question of general productivity we are asking the following three
questions:

(1) Are some negative prefixes more productive than others? It has been claimed (Marchand
1969: 169) that un- is the strongest negative prefix in English. We are seeking, in part, to
confirm or deny this claim (see also Zimmer 1964).

(2) Are the phonological rules which determine whether a prefixed in- will be realized as
in-, im-, il- or ir- (intolerant, impossible, illogical, irrelevant) productive rules for native
speakers of English, or are they simply relics of Latin morphophonology which survive in
English borrowings from Latin? Put another way, is impossible a Latin word which has
been borrowed whole, or is it an English word formed productively by native English
speakers through the application of internalized rules?

When we consider the data below, we realize that we are dealing here with a skewed
pattern of apparently unpredictable variants. Such a pattern makes the task of writing a
simple rule of nasal assimilation more difficult than it would first appear:

unmotivated *ummotivated immaterial *inmaterial
unlucky *ullucky : illogical *inlogical
enlarge *ellarge

unreasonable *urreasonable ; irreverent *inreverent
unpopular *umpopular : impossible *inpossible
inroad *irroad

The factors determining whether the nasal of the prefix will assimilate seems to be sensitive
to whether or not the -n- in the prefix is derived from Germanic un- or Latin in-. Syn-
chronically this might be most easily captured by putting the vowel of the prefix (i-) or
(u-) in the environment of the rule, to read something like “n becomes / when followed by
land preceded by 7," or in more current format, as follows:

a cor | acor

ant ant

[+ nas]> à nas Ls n
6 cont dcont

This rule restricts the assimilation to nasals which are preceded by /i/. It works provided
that inroad is analyzed as a commound with # rather than a sequence of morphemes joined
by +. (Special thanks to Carol Anderson for her help with the details of this rule.)

Processes like anticipatory assimilation of nasals are often referred to as ‘natural rules’
because they involve articulatory ease resulting from the transfer of feature values from
one segment to another. But with negative prefixes the naturalness of the rule seems to be
determined by something that should play no part in the rule, namely the preceding vowel.
This leads us to ask whether the rules of nasal assimilation in English are productive rules
or rules which are only apparently productive because of the vast numbers of borrowings
from Latin and French which contain the negative prefixes.

(3) As a sequel to the issues raised in (2), we ask the following question: are native speakers
of English intuitively aware of a distinction between Latinate (or perhaps simply foreign)
vocabulary and native English vocabulary? Do they feel that possible, tolerant, operative
and logical, for example, are fundamentally different from kind, good, fat, and thin? If
speakers can recover such knowledge about the etymology of words, how is it reflected:
by their choice of prefixes; of suffixes; by the segmental phonology of the word; the
number of syllables; the stress pattern? The possibility that suffixes mark words as Latinate
or English, and that this determines the choice of the prefix un- or in- merits closer inspec-
tion. If it is the case that speakers somehow ‘look ahead’ to the suffix before attaching the
prefix, this might account for some of Jespersen’s interesting doublets (1942: 468) like
unjust: injustice; uncertain: incertainty; unable: inability, where the Germanic un- occurs
on the bare adjectival form and the Latinate in- on the derivationally complex form with
the Latin suffix. It would also help to explain why some Latinate words with native Ger-
manic verbal suffixes favor un- prefixation, while their Latin counterparts favor in-: cf.
unanimated: inanimate; unceasing: incessant; ungrateful: ingratitude; undigested: indi-
gestible. But this solution is surely too simple: cf. unresponsive, unromantic, unpopular,
etc.

According to Marchand (1969: 168), the Latinate negative prefix in- can be considered an
independent formative in English by about 1500 A.D. This results, of course, from the

huge number of borrowings with negative prefixes from Latin and French from the earliest
times in Old English, and systematically from the 14th to the 19th centuries. Marchand
claims (1969: 169) that un- has always been stronger than in-, and is ousting it more and
more. In Early Modern English, in- could be prefixed to almost any adjective with a Latin
or French basis, and always according to the Latin rules governing assimilation. But such
words as inceremonious, incertain, inchangeable, ingrateful, incomfortable, impleasing and
impopular have all given way to un- forms. Yet there are also cases where former un- words
have given way to their earlier etymological negative, viz. uncredible, undubitable, unef-
fable, unexcusable, unexpert, unfirm, unformal, unglorious, unperfect, unpiteous, un-
possible. I am unaware of any examples in English of an etymologically native English
word which is prefixed with in- in any of its purportedly productive forms (cf. Zimmer
1964: 29). Thus, while it is true that some former un- words have gone over to in-, this has
only happened with words of Latin origin. /n- (im-, il-, ir-) is never prefixed to native Eng-
lish words to mean ‘not’ cf. *imborn ‘not born’; *irrocky ‘not rocky,’ *illoved ‘not loved,’
etc. If in- is indeed an independent formative, why is it only found with Latinate words
and not with etymologically English words?

2. The Experiment

We began our investigation of English suffixation and prefixation by studying negative
prefixes for one simple reason: while speakers have great difficulty assigning meaning to
prefixes and suffixes generally, they are consistent in assigning negative prefixes attached
to adjectives the meaning ‘not + adj.’.

Several experiments have been designed which I believe will shed some light on the three
questions above. I must stress again the preliminary nature of the results, but I hope
nonetheless that they will set the basic pattern of investigation in the open. I will report
here only on the results of the Antonym Production Survey (APS), which is a production
test designed to force subjects to generate what they consider to be the most acceptable
opposites to the words they are given.? All subjects were sophomore-level undergraduates
at The Pennsylvania State University.

The APS utilizes a data list which consists of sixty pseudo-words, thirty of which are
Latinate and thirty of which are English (Anglic)? in character and appearance. The words
were controlled for various phonological and morphological features. The Latinate words
were composed of segments, phonotactic patterns and suffixes which made them appear to
be unfamiliar Latin words. The same was done for the Anglic list, though it must be ad-
mitted that it is an infinitely easier task to devise pseudo-Latin words than pseudo-English
ones. The conditions for which we controlled in the design of the words were the following:

(1) The number of syllables

While at first we tried to match Anglic-Latinate words by number of syllables, it became
quickly apparent that the Latinate words simply had to be longer in order to appear La-
tinate. The differences were kept to a minimum by limiting the Latinate words to one
syllable more than the corresponding Anglic word (e.g. Lat. limoral (3): Eng. lurmy (2)).

(2) The suffixes
Latin adjectival suffixes are generally longer and have a more complicated internal struc-
ture than their English counterparts. Such Latin suffixes are also quite plentiful in English,

and are easily attached to pseudo-roots. The same procedure of suffix attachment was
carried out for pseudo-English adjectival suffixes and roots, but the results are less impres-
sive. The number of native English adjectival suffixes in the language is quite small, and at
least on finds it way into the system via the participial ending of the weak verbs in -ed.

(3) The initial segments

While the choice of un- or in- does not appear to be conditioned by the initial phoneme of
the root, the exact form of in- (in-, im-, ir-, or il-) most certainly is. Therefore the initial
segment of each word had to be carefully prepared. In all, we arrived at the following
distribution among our thirty pseudo-Latin words. The pseudo-English (Anglic) group has
basically the same distribution with some minor but necessary variations to conform to
native phonotactics:

a) Six words beginning with a labial (m or p).

b) Six words beginning with /.

c) Six words beginning with r.

d) Twelve words beginning with either a vowel or a ‘neutral’ consonant; i.e. neither m, p, l
or r, all of which might trigger assimilation.

The lists are:

Latinate

abfunious retropulous miniculous monarial retulous
lanurous sempient nexibular prosilient exilar
medable conable profinous repulient locanous
addorsent treculant repucent lobunious robulous
neculous effucent entulient resular fissular
picculent libular limoral exconal logenic
Anglic

appenish rotchity mitchity mofitted rimbled
lattished sontful neshorly plinkity finkled
miskful eppish prafilled rubaded loofish
arnful trookish relshy lamoted ruckful
nellful esheled eskity relby kibful
pelky liggish lonky ertful lurmy

3. The Antonym Production Survey

As mentioned, this test is designed to force subjects to generate opposites by prefixation.
Fach one of fifty-one subjects was given the following set of instructions, followed by the
thirty Latin and thirty English pseudo-words. (Note that in the instructions to this test no
examples of negative prefixes are provided.)

Instructions to the test:

In English, new words can be created from old words through the use of PRE-
FIXES. For example, re- (re-apply), pre- (pre-arrange), post- (post-date) and sub-
(sub-oceanic) are prefixes which are used to alter the meaning of some base word.

Notice that each prefix in these examples has a specific meaning: re- means ‘again’,

pre- means ‘before’, and so on, so that rearrange means ‘arrange again’, prearrange
means 'arrange beforehand'.

There are also prefixes in English that give words opposite meanings. These are
called NEGATIVE prefixes because they mean ‘not’. We are interested in finding
out which negative prefixes speakers of English actually use. To help us do this,
would you please give us the opposites of the words using the negative prefix
which sounds best to you for each word. These words happen to be pseudo-words,
but most people have no trouble deciding what the best opposite word should be.
Of course, there are no right or wrong answers, so just use your intuition about

what is best.
(Anglic)
Word Opposite
1. appenish
2. lattished
3. miskful
(Latinate)

1. abfunious
2. lanurous
3. addorsent

The results of this test are quite interesting. They are presented below in both tabular and
graphic form, with some following discussion. Listed at the top of each category is the
initial phoneme of the root according to the conditions mentioned earlier.

Table 1
Neutral

Prefix Vowel Consonant || Labial 1 r Total

PL|PE |PL| PE |PL|PE |PL|PE |PLl| PE | PL| PE
UN 115, 127 120} 186 j| 132| 161 | 116; 160 87| 145 | 570| 779
IN 57| 36 50| 16 39| 35 39| 29 86| 53 | 271; 169
(correctly
assimilated)
NON 74| 60 56| 26 70| 50 85| 52 65| 36 | 350| 224
DIS 30| 65 37| 54 29| 32 37| 38 36| 46 | 169| 235
IN 4| 7 2| 0 10) 4 6 8 12] 3 34| 15
(incorrectly
assimilated)

OTHER 17) 19 30 32 | 21-19 23| 17 21) 27 } 118) 114

P-L = Pseudo-Latinate; P-E = Pseudo-English

Figure 1: Percentage production of negative prefixes

70%

60%
50%

40%

in (correct)

Percent of total responses

10%
0%
vowel neutral labial l r
consonant (m/p)
INITIAL PHONEME

Figure 2: Relative markedness of prefixes for foreignness

ANGLIC
20%

10%

LATINATE

vowel neutral labial 1 r
consonant (m/p)

INITIAL PHONEME

Difference in Production for Anglic and Latinate Roots

From the tabular display and the graphic representations we can discern the following:

(1) Un- accounts for 44.3% of the total responses.

(2) Non- is the second choice, with 18.8% of the total.

(3) In- is third with a total of 16% of the responses; however, when we compare correctly
assimilated in- (e.g. irrobulous) to incorrectly assimilated in- (e.g. ilrobulous), we see
that nearly 90% of the in- choices are correctly assimilated.

(4) Dis- is the fourth strongest choice, with 13.3% of the total.

(5) All others (mis-, a-, no-, etc.) account for only 7.6% of the total.

Compare Figure 1 for graphic representation of the results mentioned in (1)—(5).

Let us look now at the distribution of choices according to the ‘etymology’ of the root.

For graphic representation, see Figure 2.

(6) 61.6% of all in- responses were to Latinate roots; while the remaining 38.4% were to
Anglic roots.

(7) 42.3% of the un- responses were to Latinate roots; while 57.7% were to Anglic roots.

(8) 61% of the non- responses were to Latinate roots, while 34% were to Anglic.

(9) 41.8% of the dis- responses were to Latinate roots, while a surprising 58.2% of the
dis- responses were to Anglic roots.

The general conclusions to be drawn from this test are that un- is the strongest negativizing
prefix productive in English today. Jn- is definitely preferred for Latin-looking roots than
for English-looking ones, suggesting that speakers have some intuitive awareness of the
foreigness of Latinate vocabulary, though what cues they use is a matter which remains to
be decided. It also suggests that the prefixation phonology of in- is productive in English.
The general selection of non- as a second choice is interesting, though once again it is much
stronger for Latinate roots than for Anglic ones.

A mild surprise in the data is the preference of dis- in Anglic roots to dis- in Latinate roots.
This results no doubt from the synchronically neutral look and feel of dis- which make its
Latinate origins opaque. Unlike a-, in-, or non-, dis- has been completely incorporated into
the English lexicon.

Thus the high productivity of un- and the low productivity of in- for Anglic roots seems to
confirm Zimmer’s results and Marchand’s claims about un-/in- productivity.

4. Conclusions and Speculations

It is clear that un- is the preferred negativizing prefix in modern English, and that non- and
in- are its closest rivals. It is also clear that speakers have very strong feelings about the
relative nativeness of words, and that these feelings guide them in their selection of a nega-
tive prefix. What is most interesting is the fact that if in- is chosen, it will almost always be
correctly used. Thus in- seems to be morphologically restricted, but the rules governing its
use are phonologically productive. Why these rules have not spread to un- is an interesting,
if unanswerable question, as is the converse, viz. why hasn’t in- leveled out its variants to
match the pattern of un-? Perhaps the in- words really don’t have two morphemes, and
words like illogical, impossible and irreverent are stored separately in lexical memory from
their positive counterparts logical, possible and reverent. The answer to these questions lies
in the future.

44
Notes

1 As of this date (May, 1983), two experiments have actually been run, though the complexities of
tallying the results leave me with only one to report on in detail here. Two more are being prepared,
and will be administered in the coming months. The present paper represents only a fragment of a
much larger study being conducted by David Palermo, Victor Broderick and the present author
which is designed to test English prefixation and suffixation in general. I hereby acknowledge my
indebtedness to these two collaborators, but free them from blame for whatever faults lie in the
present paper. Broderick prepared the graphs and helped me with the statistical calculations. I would
also like to thank Carol Anderson and Donka Farkas for their stimulating comments.

2 Un-, in-, non- and dis- are not, of course, the only prefixes in English with negative force. In fact,
English abounds in such prefixes: a- (an-) (amoral, anorexic); n- (none, neither); no- (no-win situa-
tion); mis (misbegotten). 'These are all quite marginal and highly marked lexically in comparison with
un-, in-, non- and even dis-, which is itself of marginal productivity.

3 We have also administered a Multiple Choice Antonym Production Survey in which subjects are
asked to list their first and second choices from a register of six potential opposites. This test, which
is much more complicated than the Antonym Production Survey, will be discussed in a future paper
along with the results of other experiments. Preliminary analysis of the data suggests a confirmation
of the results found in the Antonym Production Survey.

^ The entire question becomes more complicated when we take into account such matters as the
educational levels of respondents, their degree of foreign language awareness, their attudes towards
‘learned’ vocabulary and other factors which influence speakers’ judgments and choices.

5 The term *Anglic' is used to designate pseudo-words which are supposed to appear to be of native
English origin. ‘Anglic’ is opposed to ‘Latinate’.

References

Jespersen, Otto, 1942. A modern English grammar on historical principles (Part VI-Morphology)
(Copenhagen: Munksgaard).

Marchand, Hans, 1969. The categories and types of present-day English word-formation (2nd ed.) (Miin-
chen: Beck).

Zimmer, Karl E., 1964. Affixal negation in English and other languages (Supplement to Word, Mono-
graph 5) (New York).

Indo-European neuters in -

Robert S.P. Beekes (University of Leiden)

0. The aim of this article is (1) to present the evidence for neuter i-stems with full inflec-
tion and (2) to show that there is no evidence for i/n-stems nor (3) for neuters with -i in
the nominative only (and zero, i.e. root noun, in the other cases).

1. Neuter i-stems (with full inflection)

1.0. Handbooks often speak of neuters in -i, but mostly this refers to adjectives. It appears
that there are hardly any neuter i-stems in the Indo-European languages, but that there
were in Proto-Indo-European action nouns which were very important in the history of the
verb (section 1.9.). We shall present the evidence of the separate languages. This survey
does not claim to be exhaustive in details except where indicated.

1.1. Indo-Iranian
1.1.1. Sanskrit

The four i/n-neuters are mentioned everywhere; they will be discussed in the next chapter.
The handbooks further mention härdi, which belongs to section 3.2., and väri, which re-
places older vär (Wackernagel 1954: 291ff, spec. $ 190c; 1957: 131, 145, 160).

I checked the suffixes in -i indicated in the reverse index of Grassmann's dictionary (1719
22) and Wackernagel (1954). I found no neuter.

Burrow (1955: 175-177) mentions sami and srkvi, which 1 cannot find (for the latter cf.
Mayrhofer 1976: 554 sräkva-). That sdci ‘with’ is an old neuter is uncertain. Burrow thinks
that some forms in -ya continue older neuters in -i: nabhya- n. beside nabhi- f. (where the
long vowel rather points to an original root noun), madhya- (which is certainly of Proto-
Indo-European date). On kravya- see section 3.1.

Further he believes that arcis-, rocis-, socis- were neuters in -i. But roci-, which he adduces
as evidence, “ist jung (Pur., Hariv.) und nicht ganz sicher (PW VI 441)”, Mayrhofer (1976:
76). See on these words Wackernagel (1954: 365f.).

1.1.2. Iranian

In Old Persian only dipi- ‘writing’ is in one place supposed to be neuter, but others take it
here too as feminine (Brandenstein-Mayrhofer (1964: 116)). The word is a loan from Ela-
mite.

For Avestan neither Bartholomae (1895—1904: §§ 189, 406) nor Reichelt (1967: $$ 303f,
363) mentions a neuter. In the Gatha’s proper and the Yasna Haptanhaiti I counted 129
neuters (and 17 doubtful cases), but no i-stem. Of the reverse index, Bartholomae (1961)
I checked the gender of the forms in -i not ending in -ti. I counted some 130 words. Two are
listed as neuter. z&iri ‘käsig gewordene Milch, Molke’ has been connected with Gr. turds
‘cheese’ (because of Myc. turoy /turjos/ from *turio-?). There is no good etymology. I
don’t see on what basis the word is considered a neuter (only tuiringm N 66 and 67). The
word may be non Indo-European, or it could be the neuter of an adjective.

The other word is tayuiri- ‘bread’. Again I don’t see how we know that the word (only
tayüiringm V 16.7) is neuter. It could be an adjective, as is xXaudringm, ibid. (Nor do I
understand why Bartholomae refers from the one word to the other and to tarsu-.) The
word has no etymology.

The conclusion is that I find only two words given as neuter, both of which are in fact of
unknown gender and which have no good etymology.

1.2. Armenian gives no information as the gender distinction has been given up.

13. Balto-Slavic
1.3.1. Baltic

In Lithuanian and Latvian the neuter has disappeared. Old Prussian, which still has neuters,
has no evidence for an i-stem (Trautmann 1910: 235ff). (mary ‘hab’ Voc. is an &-stem.)

1.3.2. Slavic transfered the i-neuters (if there were any) to masculines or feminines (Meillet
1934: 417).

1.4. Tocharian has neuter forms only in the pronouns.

1.5. Hittite

Neuter i-stems are frequent. Brosmann (1978) counted 87 of them. He points out that 15%
of them are loanwords (because there are good etyma or because they have Hurrian end-
ings). "In view of the large foreign element in Hittite, the scant attestation of Hattic and
Hurrian and the lack of an etymology for a majority of the neuter i-stems, one can be fairly
confident that the actual proportion of such loanwords was considerably greater. . . At the
other extreme, evidence concerning words of known original gender inherited from Proto-
Indo-European is largely non-existent.” Only the suffix -asti is clearly Indo-European. dalu-
gasti ‘length’ and pargasti ‘height are neuters, palhasti is variable. As word with this suffix
are always feminine in Slavic, it is supposed (Kronasser 1966: 209) that they became
neuter in Hittite. But Kronasser did not give an argument why he rejected Pedersen’s view
(1938: 35) that the words were originally neuters. This view seems more probable, first,
because Slavic does not have any i-stem neuters at all, and secondly, because a transition
from neuter to feminine seems in general more probable than vice versa. This development
may be rather important: first, it would show that Hittite retained neuters where they
disappeared elsewhere, and secondly, it might indicate that the i-neuters, which were often
action nouns (section 1.9.), became feminines in other languages too.

Kronasser (1966: 203) writes about the primary i-stems that there are hardly any trust-
worthy Indo-European connections (“Wortgleichungen, wie bei den a-Stämmen, finden
sich darunter nicht"). Among the neuters he mentions lissi ‘liver’, for which Schindler
(Sprache 12, 1966: 77—78) proposed an etymology. He interprets Arm. leard as *lis-r-t, to
which lissi would be a variant in -i. (The root vocalism is not essential here. Nor are -i/-r
variants probable, but Arm. -ard can be analogical.) But it is possible that the Hittite word
was an old neuter in -i, though it remains possible that the -i is an Hittite addition (cf. meni
n. ‘cheek, face’). Schindler thinks an adjective, ‘the fat one’, is possible, but his connection
with Lat. Jardum etc. should be given up.

The conclusion is that a considerable increase of i-neuters was caused by loanwords (cf. on
Greek below). We should also bear in mind the notorious productivity of the i-stems in
Luwian. In Hittite there are 195 c and 87 neuter i-stems against 24 c and 17—19 neuter
u-stems. It is clear that the Proto-Indo-European situation was reversed in Hittite. Though
there are no words with an Indo-European etymology, it is quite possible that there was an
Indo-European kernel.

1.6. Greek has loanwords like sinapi, péperi (Schwyzer 1939: 462). Chantraine (1933:
114) adds iskhi* osphus Hesychius. The gloss may be “une graphie tardive, ou un simple
faute pour iskhion (ainsi Latte)", Chantraine (1968— 1980). Chantraine himself connected
the word with iksus, which proves non-Indo-European origin for Furnée (1972: 393). Non-
Indo-European origin is anyhow probable. dlphi (see section 2.0.) and meli (see section
2.0.) are f-stems. thémis is “vereinzelt und sekundär neutrum” (Frisk 1972 s.v.). The
theory (e.g. Benveniste 1935: 34) that the word was originally a neuter *themi, -itos
cannot be proven. It rests partly on the idea that the Sanskrit neuters in -is were originally
i-stems, which is most probably incorrect. If *themi, -itos were correct, it could also have
been "themit. In the same way könis f. ‘dust’ would have been a neuter because of Lat.
cinis, -eris, under the assumption of an s-stem *konis-, *kenis- and behind that a neuter
*koni. If this is correct, it lies far back. Indirect evidence would be ostéon and ósse, which
are discussed below.

The reverse index of Buck-Petersen (1945: 14) states clearly that there are no inherited
neuter i-stems.

In Latin I find only two or three words, mare, rete ‘net’ and ?ile. rete has no etymology
and cannot be used as evidence for a Proto-Indo-European neuter. The first word has been
considered to be of non-Indo-European origin by Nehring (1959: 122) because of the a/o
interchange: The neuter plural ilia ‘flancs, parties latérales du ventre” has in the singular
Tlium, ileum, ile, which suggest that ile < *ili was the oldest form. It has been connected
with (Ma: udpia (codd. 6copa!) yuvaeia; Mov: TÒò Tüc Yuvaòs Eynßarov noi. kat
Kooputov ‘yuvaiKetov mapa Kao. (Cf. Pokorny 1959: 499). Indo-European origin is far
from certain.

Germanic i-neuters are found in Gothic and Old Icelandic (marisaiws in Gothic contains

the word for ‘sea’). Only one word is found in more languages, *mari ‘sea’: OHG meri

(also masculine); in the other languages it has become masculine (i-stem) or feminine

(stem in -in). Further we find:

OHG bini ‘bee’; as the -n- is a Germanic addition, the i-neuter is a Germanic innovation.
quiti, queti 'Ausspruch'.

OS bini.
urlagi ‘war’; the word is a Germanic creation; see De Vries (1971: s.v. oorlog).
hals-meni ‘necklace’; OHG menni, OE mene, Olc. men show that it was a io-stem.
land-skepi, friund-skepi etc.; in any case a Germanic creation; see De Vries (1971:
s.v. schap 2).
OE spere ‘spear’ belongs to Germ. *speru-/sparu-.

There is no evidence for an Indo-European form. On ‘sea’ see under Latin.

1.7. Celtic British gave up the neuter.

Old Irish.
Thurneysen gives seven i-neuters in his Grammar (1946: 191). Only two or three have an
etymology.

guin ‘wounding’ must be *g“honi. (Cf. section 1.9.)

muir ‘sea? < *mori.

búaid ‘victory’ has been compared with Boudicca and Germ. Baudi-hillia ‘Sieges-
kämpferin’, which would point to *bhoudhi- (Pokorny 1959: 163).

cuirm ‘ale’ belongs with Gaul. Korma, kourmi. Further connections are (extremely)
uncertain.

druimm ‘back’ is supposed to be a loan from Welsh (Pokorny 1959: 1075).

graig ‘horses’, gen. grega. Not a loan from Lat. grex according to Pokorny (1959:
382).

richiss ‘live coals’ (see Thurneysen (1946: 191) for the gender) has been connected
with Lith. f. pl. piřkšnys, but Vendryes (1959: R 29) notes that even the Celtic
form cannot be reconstructed.

Note that cuirm and druimm have the o-vocalism expected in Proto-Indo-European (as
least in one type, see section 1.9.).

1.8. Conclusion

There are two kinds of danger with articles like this. One is that the author cannot resist
the temptation to deny every form of evidence. The second is that the nuances disappear
when the results are cited (“B. has shown that Proto-Indo-European had no i-stem neu-
ters"). Therefore I shall try to be very clear in the conclusion.

Evidence for neuters can be expected from Indo-Iranian, Old Prussian, (Slavic), Hittite,
Greek, Latin, Germanic and Celtic. There is no positive direct evidence in Old Prussian (and
Slavic). There are hardly any i-neuters in Indo-Iranian, Greek, Latin and Germanic, and
probably none that is of Proto-Indo-European date. Positive, direct evidence is found in
Hittite and Old Irish. In Hittite there was a large non-Indo-European influx. Of Proto-Indo-
European date could be the words in -asti and lissi ‘liver’. Whether there are more inherited
words should be investigated. In Old Irish guin ‘wounding’ will be old, for other words this
is not certain. Doubtful is *mori ‘sea’.

As we find i-neuters in two ‘extreme’ languages (Hittite in time, Old Irish in position), it is
possible that they retained i-neuters from Proto-Indo-European and that the category was
enlarged by loans and new formations but lost the old forms. On the other hand, in Indo-
Iranian and Greek they seem to have been lost very early. A certainly old type is discussed
in the next section.

1.9. Recently it has been supposed that there were Proto-Indo-European neuters of the
type CoCi that were very important for the development of the verbal system.

Burrow (1955: 177) thinks that the (3rd sg.) passive aorist type tari, jéni, continues a neu-
ter in 4. Kortlandt UF 86. 1981: 127) shares the idea (1981: 127). As the type is also
found in Avestan (GAv. vaci, sravi), the passive aorist is at least of Proto-Indo-Iranian date.
If Kortlandt is right in assuming that the causatives are derivatives of such verbal nouns, the
type must be very old. He also assumes (1981: 128, note 1) that such a noun was used in
the formation of the Germanic weak preterite (e.g. Gothic 2nd sg. nasides ‘you saved’ <
*nosi dhes) and of the Old Irish denominatives, type -suidigedar < *sodi sagitro (see Thur-
neysen 1946: 8 524).! This would mean that this type of i-neuter must date back to Proto-
Indo-European and lived on into the separate languages, although perhaps already as a
fixed (indeclinable) form with restricted use. No doubt Olr. guin ‘wounding’ is a remnant
of these verbal nouns (it is isolated in Irish, Thurneysen 1946: 448).

Rather speculative is the following suggestion. The neuter dual ending -ihj (Gr. össe <
*ok"ih1) might have its i from the neuters in -i. This would prove their former importance
in Proto-Indo-European. (Note that du. -/1,, pl. t5 was a not well marked system.)

2. i/n-Neuters

2.0. Four Indo-Iranian neuters have i/n inflection. On this basis an i/n-inflection for Proto-
Indo-European parallel with the r/n-stems has been postulated (the first was J. Schmidt
1889: 248ff). I do not think this is correct.

2.1. The evidence outside Indo-Iranian is very untrustworthy. What Benveniste (1935:
6—8) presents "gehórt entschieden zu den Teilen der Jagdbeute, die Benveniste aus der
vollen Jagdtasche als unrechtmässig erlegt wird ausliefern müssen” as Pedersen (1938: 17 n.
1) said of parts of “das wertvolle r/n-Wild".

Passing by Lith. vagis etc., Lat. axis etc. (and the word for ‘ear’, where both Av. uXibya and
Lith. ausis are innovations), we note that for Skt. hardi etc. the Germanic forms do not
prove a Proto-Indo-European n-stem, and retain only three words: Gr. alphi ‘barley’, Lat.
mel and Lat. sal etc.

For aiphi, normal plural dlphita, the Hesych gloss alíphata dlphita e aleura would show an
old n-stem: it would replace *d/phata. Unnecessary to say that this interpretation, based on
a gloss, is not certain enough to prove anything for Proto-Indo-European. Latte corrects
the form into *alephata ; cf. Chantraine (1968— 1980: s.v.).

Lat. mel, mellis would represent *meli-t, melnes. But the n-stem is not certain. Ernout-
Meillet (1959: s.v.) consider -In-, -Id- or “ancienne géminée populaire”. Leumann (1977:
213) follows Szemerényi's idea (KZ 75. 1958: 183 n. 1) that it is analogic to fel, fellis
‘bile’. In any case the n-stem is found nowhere else. From Hittite melit no oblique cases
are known, but the adjective meliddu- ‘sweet’ and Luw. malli, pl. mallitinzi are based upon
the form with f, so t-flection seems probable. Gr. blitto must be based on a stem form
*mlit-, which points to an old inflection *melit, mlit-és.

‘salt’

“Auf Grund von aksl. slana (‘gesalzen’), air. salann ‘Salz’, gr. hdlasin huei (Suid.) setzt
Schmidt (1889: 182) einen obliquen Stamm *sal-n- neben den nom. *sal-d oder *sal-i an,
eine Annahme, für die jedenfalls der anscheinend späte griechische Ausdruck keine Stütze
bilden kann”, says Frisk (1960: 79). The Slavic adjective is of course a no-adjective derived
from *sol- (Meillet 1934: 267), the Celtic word is supposed to continue *saleino- (Pokorny
1959: 878) and does not point to *sal-n-. The athematic inflection of Greek, Latin (the
nom. sale is secondary according to Ernout-Meillet) and Slavic (the i-stem is shown to be
secondary by *sol-no-) must be old.

The non-Indo-Iranian evidence, then, appears to be non-existent. It should also be observed
that Hittite, where we find so many r/n-stems, has no i/n-stems.

2.2. Sanskrit has four i/n-neuters: dsthi, sdkthi ‘thigh’, dadhi ‘sour milk’ ,dksi ‘eye’ (Wacker-
nagel 1957: 302—306). However, the Avestan cognate of dsthi has the following forms ac-
cording to Kellens (1974: 336—339) (younger forms in brackets, masculine/feminine forms
and a-stem forms):

singular dual plural
nom. asti
acc. as-ca, (astam) (asta-ca ) asti-ca, (astas-ca, asta-ca)
gen. asto, astas-Ca astam, (astangm)
instr. azdobis, azdibix 7 Jazdbix/

It is evident that the oldest inflection was that of a root noun, and this must be the Proto-
Indo-Iranian inflection, and, we may add confidently, the Proto-Indo-European inflection.

Avestan has one form that suggests an n-stem, astantat- YH 41.3. (First it should be recog-
nized that *ast-tat- was difficult. Note further that we would expect /astatat-/, cf. GAv.
karapo.tat- /karpatat-/. I think that -an- was taken over from a form like Skt. astanvánt-
(GAv. has astvant-), where -an- was regular before -v- If this is correct, it proves the exis-
tence of other forms in -an-.) Iranian, then, may have had n-forms too, but this cannot
invalidate the conclusion that the root-inflection was the oldest type.

Gr. astakos/ost. ‘lobster’ is considered to represent -n-ko-, but because of the vocalism this
is uncertain. The word cannot be a Greek formation, and it is hard to believe that it is of
Proto-Indo-European date. The word is probably non-Indo-European. (Cf. Beekes 1969:
51; Furnée 1972: 137, who thinks that stakhös is not old. óstrakon does not represent
*Hostrko-, or even *Hostrnko-; nor is astrágalos a Proto-Indo-European *Hstr-g-(h»Jlo-.)

Further support could be seen in Venetic, where ostinobos (in Latin script) is interpreted
as ‘ossibus’ (Lejeune 1974: 337) and explained from *ost-n-. But it could also be an ad-
jective in -tino- (id. 99). ostiiako (olim ‘ossuarium’) has turned out to be a personal name.

Sákthi has in the Rigveda further an i-stem form sakthya and the n-stem sakthani. The
Avestan cogante appears in three forms, acc. du. haxti, gen. haxt(ajya and haxta (see sec-
tion 3.4.), none of which is an n-stem. As the i-stem can be explained (e.g. from the dual),
we may suppose an original consonant-stem as for dsthi, which may be confirmed by Aaxta.

For dádhi OPr. dadan, which is an o-stem neuter, shows that there was no i/n-stem.

Du. aks? shows (cf. the accent) that the word for ‘eye’ was a consonant stem.

Compounds of all these words are from consonant stems: Skt. an-asth-d-, -sakth-d-, an-dks-
(Wackernagel 1954: 93, 108f; "aber bei uneigentlicher Bedeutung [that is in younger
forms] -aksi-, -sakthi- P. 5.4.113"). Though in itself this is not definite proof for consonant
stems, it confirms the other evidence.

2.3. The conclusion is that neither Indo-Iranian nor the other languages have evidence for
Proto-Indo-European i/n-inflection.

3. i/zero-Neuters

3.0. Is there any evidence for neuters that had -/ in the nominative but no suffix in the ob-
lique cases? We shall first look outside Indo-Iranian.

3.1. It should be remarked in advance that the distinction between i/zero neuters and
neuters with complete i-inflection is difficult. One aspect of the problem is that neuters
often occur only in the nominative-accusative singular.

I can find only one form for which -i in the nominative has been supposed (and -n- in the
oblique cases, which, however, proved, improbable, supra section 2.0.), the word for ‘salt’
(Benveniste 1935: 8). However, Lat. sale is secondary, OCS solë replaces “sol as appears
from slana ‘salted’, Gr. hali- in compounds (against háls) does not prove a neuter in -i
(though Chantraine still states this possibility, apparently because Benveniste’s authority
is still strong; cf. below).

Gr. iskhi was discussed above (section 1.6.).

We need not discuss Benveniste’s theory (1935: 75ff) of a large scale transference of i-
neuters to masculine-feminines. The least one can say about it, is that it has not been
proven. To posit, e.g. *ikri, *orni for íkrion, órnis órneon? is gratuitous. Certainly wrong
is it to take klönion (a gloss in Hesychius) as a basis for *Kloni, as klónion is clearly based
on Klónis (after words like iskhion; the gloss begins with klonion’ iskhion), which is fe-
minine, as are the cognates Skt. sroni-, Av. sraoni-, Lat. clunis, W. clun; OPr. slaunis Vo-
cabul. is not neuter, so probably feminine; Olc. hlaun was mostly given as neuter, now
mostly as feminine (as far as I know the gender cannot be ascertained for Old Icelandic; as
it is at present neuter, this gender was assumed for Old Icelandic too; this is improbable,
also because the language has no other i-stem neuters). Chantraine (1968—1980) again still
posits the possibility of a neuter. Nor is there any reason to posit a *kreuHi n. for Skt.
kravyám (d-kravi-hasta- may contain "kreuH- or *kreuHi-, but it is not necessary to posit a
neuter for it), Lith. kraijas, OPr. krawian.

3.2. In Indo-Iranian Sanskrit has AZrdi, the four i/n-stems and vari. The latter replaces
older var. This i-stem may have developed from an -i added to the nominative singular, but
this is not certain. In any case the nominative in -i was evidently a younger form. There is
no reason to suppose that vir (RV) is a secondary form (as does Wackernagel 1957: 34).

The word for ‘heart’ is now generally reconstructed as:

kerd pl. Kerd-h?
Krd-és krd-óm
etc. etc.

(Rerd-i? )

It was recently discussed by Szemerényi (1970), with whom I agree on most points. (I am
not convinced that the nominative was *kér in Proto-Indo-European. I do not believe in a
development *kerd > *kerr > *ker.) He has definitely refuted the idea, repeated over and
over again in all handbooks, that there was an old i-stem. In fact many languages prove an
old consonant stem (Lith. Sirdî etc. prove that širdis is recent; and even if we would not
have that evidence, Xirdis cannot be used as proof for an old i-stem, as is done so often;
recently by Brosman (JIES 6. 1978: 98), though he knows Szemerényi's article), and the
forms with -i- can be explained as younger formations (Gr. kardía has a suffix found in
other words for parts of the body; Olr cride < *krdiom has nothing surprising; Hitt. kar-
dias does not prove an old i-stem and there is the regular genitive kardas; see also below on
kardias). The Armenian i-stem is not sufficient to prove Proto-Indo-European origin; Sze-
merényi (1970: 526) reminds that Armenian has more often unoriginal -i-stem forms (e.g.
from otn ‘foot’).

It is most improbable that an i-stem existed by the side of this root formation. In that case
most languages would have chosen the /-inflection, which does not present the difficulties
of the root noun. A full -inflection has been assumed especially to account for Skt. hydavam,
LAv. zarodaem. (The same explanation is often given for Gr. ostéon.) Not only is a full
i-inflection most improbable, even if it had existed it would probably not have given, with
thematization, the Ilr. -zya-suffix, as there is no evidence for such a development (Wacker-
nagel 1954: 213—215; there is evidence for -ya- from hysterodynamic i-stems, ibid. 807f).

I think the explanation of the forms in -ayam is the following. “Auch sonst ist im Veda der
NAVSg. n. mancher Stämme durch andere Bildungen vermieden” observes Wackernagel
1957: 32) and gives hrdayam,as an example. I think this is the explanation: the form re-
places the difficult nominative (*gherd > *har). The form is at least Proto-Indo-Iranian,
and as its formation in Indo-Iranian is not understandable (Wackernagel 1954: 213), it
seems to be of Proto-Indo-European date. The comparison with Gr. ostéon and Lat. hor-
deum seems quite apt: in all instances the old form was a root noun that presented diffi-
culties (specially in the nominative). The explanation of the suffix must be that given by
Risch (1974: 132) for osteon: “eigentlich ‘Knöchernes’”, i.e. -eios was the suffix meaning
‘belonging to’, well known from the adjectives indicating materials. The substantivized
form of this adjective would get almost the same meaning as the noun from which it was
derived. This explanation is also given by Szemerényi (1970: 525), who gives the curious
parallels that Lat. cuprum, fagus were in the Romance languages replaced by cupreum,
fageus. | wonder whether the Hittite genitive kardias is not simply this adjective, krdeios
‘belonging to the heart, of the heart’. Compare the Luwian adjectives in -assi- replacing the
genitive.? (A secondary formation on the basis of the dative and instrumental (Kartit)
seems improbable to me: these starting points for i-inflection were present with all con-
sonant stems; and why would only the genitive shift to the i-stems?) I would even consider
the possibility that hastiias < *HostH-eios was the origin for the rather enigmatic inflec-
tion hastai, hastiias. (Once the adjective was incorporated as a genitive, apparently an
i-stem form, analogic spread of an i-stem was only to be expected.)

The conclusion is that the -i of hardi is not cognate with the i's of the words for ‘heart’ in
other languages nor with that of hrdayam. Proto-Indo-Iranian inherited *kérd *krdés with
an adjective *krdeios of which the neuter *krdeiom could be used in the place of *kerd.
The question remains from where the -i came.

Szemerényi (1970: 526) assumes that the - developed from a prop vowel before words
with initial consonant. He gives parallels of languages that have such (final) vowels. But
here the difficulty of this solution becomes clear: some languages have such phenomena,
but then rather frequently, others do not have it. And as Sanskrit has no evidence for such
a development, this explanation cannot be accepted.

I can think of only one source, which to my mind is evident: härdi hrdas was formed after
dsthi *asthds. Essential is that the -i of asthi can be explained. Note that Adrdi and dsthi
belong to the same semantic sphere.

3.3. Remains the group asthi *asthas. The -i can hardly date from Proto-Indo-European, as
there is no evidence for i/zero neuters elsewhere. The other languages give no support for
a nominative in -i in this case. Hitt. hastai, -üas does not have -i. Gr. ostéon has been ex-
plained above (from an eio-adjective). Arm. oskr < *ostuer and Lat. os, ossis have no trace
of an -i. When the -i is of Indian origin, it cannot have been taken from härdi, where it must
be explained itself, and there were no other neuters in -i. The -i must have originated in the
word itself. The explanation has been given by Hamp (1953: 137—141). It may be well to
draw attention to his explanation, as it is not mentioned in Mayrhofer's dictionary (1953:
67, 553; 1976: 637f). The -th- shows that the word had a laryngeal, and *HóstH, HostH-és
gave “ist, *asthds, with generalization of the -th- in Sanskrit dsthi (cf. pdnthah). I arrived
at the same conclusion independently. (Szemerényi (1970: 526, n. 61) does not accept the
laryngeal, but without any argument. Kellens (1974: 336) says that Gr. ostéon disproves a
laryngeal, which is not true. Note that Hittite and Greek would have got *hasta and *oste/
a/o respectively in the nominative.)

The development to *asti was Proto-Indo-Iranian (Beekes, IIJ 23. 1981: 275—287). hürdi
occurs also in Dard- and Kafir languages (Mayrhofer 1976: 605), which means that härdi
probably existed already in Proto-Indo-Iranian. If härdi got its -i from *asti, this must be
of Proto-Indo-Iranian date, too. (Cf. below on possible Iranian *dadi, section 3.5.).

Avestan nom. sg. as-ca instead of *asti then presents a problem. Hamp here makes a sug-
gestion which is apt to discredit his theory: that Av. “ast represents “Host, supposing that
-H was a collective suffix. This is unacceptable. First, Sanskrit and Avestan must be derived
from the same form. (And if we assume *Host for Sanskrit, we cannot explain its -i.)
Secondly, a collective suffix for (the singular of) ‘bone’ is quite improbable. We must look
for another explanation. (Note that those who assume Proto-Indo-European *HostHi have
the same problem.) I see three possible explanations (which do not exclude one another):

(1) In derivations (astvant-), in compounds and before clitics (as-ca!) the -H was not
vocalized in Iranian (as normally in interior syllable).

(2) The inflection *asti, astas could lead to a new nominative *ast (a tendency strengthened
by the development in 1).

(3) The plural, which was rather frequent given the meaning of the word, was asti (<
*HostHH). To avoid the homonymy with the plural the singular could have been
reshaped to "ast. | think, then, that as/t) replaces *asti.

Whether */fost// originally contained a suffix -tH cannot be made out. The comparison
with sdkthi suggests it, but this word could have been assimilated to asthi.

3.4. Skt. sákthi *thigh' may have -i « -H just as dsthi. It is of course possible that it got its
-th- and/or -i from dsthi, but there is no necessity to assume this. In any case stipti- ‘shoul-
der’ did not take over -th-. The n- and i-inflection (du. sakthyä, probably for original
*sakthi) point to a root noun.

We do not know what the nominative singular was in Avestan. The gen. pl. haxt/a)yä
/haxtiah/ is an i-stem. This form may be analogic after the nominative dual in -/, but it
might also be analogic after the nominative singular (cf. GAv. janayo ‘women’ from jani- <
*gWenH). Of the form haxta (F 3 g ‘haxt’) we do not know which case it was. I see four
possibilities:

(1) An n-stem nominative is improbable.

(2) A locative (singular) of an i-stem is not very probable for this word.

(3) A good possibility seems to be that it is the instrumental singular of a consonant stem.
(4) Lastly, it might be the nominative dual of an a-stem (cf. the a-stem forms of ast-).

GAv. haxt(i)- does not agree with sakthi: we expect *sakth- > *haxa6-. Again it is possible
that the -f- was taken over from Av. ast- (with regular sth > st). But in the nominative
*haxti is regular from *sakti < *saktH. Compare for the non-generalization of -fh- in Av.
pantà pantànam against Skt. pánthah (and, more generally, aogo aojah- against Skt. Üjas).
However, in this case we find -f- in the oblique cases. This can be explained if the i-stem
originating in the nominative was generalized. Lastly, the -t- of the nominative might have
spread to the other cases after the example of *asti.

3.5. With dádhi, dadhnds have been connected OPr. dadan ‘milk’ Voc. and Alb. djathé.
dadan is probably an o-stem. (The only neuter n-stem in Old Prussian is semen; cf. also
Trautmann (1910: 8 157b) on wundan.) The Albanian word represents *dedh-. This gives
a problem for OPr. -a-. Therefore, Szemerényi (KZ 73. 1958: 81 n. 5) suggested that dadan
is a loan from a Germanic word for milk, comparing it with Gothic daddjan ‘suckle’. I
think Toporov (1975: 284—286) is right in objecting: (1) that we have no other evidence
for this Germanic word; (2) that the Slavs took their word for milk from Germanic, but
that was *melko-, not *dada-; (3) that dada- occurs in Prussian geographical names. The
last point seems not certain to me, the first two, however, make the suggestion improbable.

I see three possibilities for OPr. -a-:

(1) There was assimilation *deda- > dada-, but this is ad hoc.

(2) In the vocabulary a often is found for older e (Trautmann 1910: 8 lic). It is a pecular-
ity of that dialect of Old Prussian.

(3) The original inflection of the word had o/e-ablaut. (A fourth possibility is that Alb. *e
is not original.)

The difficulty, then, is not strong enough to deny the at first sight evident relation of
dadan with dádhi. I think there are two more arguments in favour. We now know that (if
the word was inherited) it must have had -dh-, for Proto-Indo-European -d- would have
made the preceding vowel long according to Winter's Law. Further, another gloss (690) has
ructan dadan ‘sour milk’, which shows that this word could be used in a phrase with
exactly the same meaning as dddhi.

Therefore the cognates must be accepted. This means that the word did not have n-stem
forms originally. But an original i-inflection as is usually assumed is equally impossible.

In Sanskrit the i-inflection might have been replaced by the i/n-inflection, (though I think
such a development is improbable), but the absence of an i-stem in Albanian and Old Prus-
sian cannot be explained.

The word is mostly derived from *dheh,-, “dhehi- ‘suckle’ with reduplication. But a pre-
form *dhe-dhhyi or *dhe-dhh-i is impossible, because this would have given a full i-inflec-
tion. Hamp therefore rejected the connection with this root altogether, but this is not
probable. See the parallels given by Toporov (1975). The only possibility left is that the
root had the i-less form and no suffix -/ (neither throughout nor in the nominative only, for
this would certainly have led to i-inflection in some of the languages). We thus arrive at the
reconstruction *dhé-dhh,, gen. *dhedhhj-és, which gives directly (Proto-Indo-Iranian)
*dadhi, *dadhás. The fact that the laryngeal of the root explains the i/zero inflection of
Sanskrit directly can hardly be a coincidence. The reduplicated root noun might surprise.
Kortlandt points out to me that the word for ‘beaver’ may have had this formation. We
find *bhebhro- and *bhebhru- (Avestan has an i-stem) side by side, and this suggests that
the word was simply *bhebhr originally.

Interesting is Szemerényi’s theory that Hungarian fej ‘milk’ derives from an Ir. *dadi (apud
Altheim 1951: 77f). It would point to an Iranian form in -i.

3.6. About dksi we must be short. We have seen that aksi *aksas probably was the original
inflection. The structure of the word is totally unclear, other languages pointing to simple
*Hok"-. It has been assumed that the -i originated from the (frequent) dual aksi. I would
expect a complete i-inflection in that case, but it is not impossible. On the other hand, a
form in -H cannot be ruled out either.

3.7. The results for Proto-Indo-European may be given in this table:

neuters in -i i/n i/zero

*ToRi: llr. pass. aor. none none
Germ. pret. (*HostH | PM *asti
Olr. denomin. *HostH-és 2 *asthds)

Skt.— Gr.—

Av. —! Lat. —2

OPr. — Gm.

Hitt. + Olr. +

1) two forms given uncertain; 2) two or three words; 3) recent forms.

Notes

1 Proterodynamic *k“ori, *k“r-ei-, younger *k“orei-, would explain the zero and o-vocalism of the eie-

verbs.

Hitt. haran- shows that the word was originally an n-stem. (The nominative haras continues *Horo <
*Horon, as Kammenhuber (1969: 289) has shown, (note that Hittite and Palaic write haras with long
a); not mentioned by Tischler (1977) who refers to the theory that the nominative had *-ans, which
would have given -anz.) Benveniste’s *or- *oren- (1935: 24) thus loses its basis. (Tischler wrongly
cites *oren-, *orn-.) Gr. orneon may have the same suffix -eio- as has ostéon (see the text, below).
There is no reason why Gr. örnis would be an old i-stem. Benveniste made the mistake to assume
that -eio- requires the former existence of an i-stem.

Comparable is Szemerényi's explanation of Gr. gunaikos as an original adjective (A/ON 2. 1960: 13—
30).

References

Altheim, Franz, 1951. Geschichte der lateinischen Sprache (Frankfurt: Klostermann).

Bartholomae, Christian, 1961. A/tiranisches Wôrterbuch (Berlin: de Gruyter).

Bartholomae, Christian, 1895-1904. Grundriß der iranischen Philologie (Straßburg: Trübner).

Beekes, Robert S.P., 1969. The Development of the Proto-Indo-European Laryngeals in Greek (The
Hague: Mouton).

Benveniste, Emile, 1935. Origines de la formation des noms en indo-européen (Paris: Maisonneuve).

Brandenstein, W. — Mayrhofer, M., 1964. Handbuch des Altpersischen (Wiesbaden: Harrassowitz).

Buck, Carl Darlins — Petersen, Walter, 1945. A reverse dictionary of Greek nouns and adjectives (Chica-
go: University of Chicago Press).

Burrow, Thomas, 1955. The Sanskrit Language (London: Faber & Faber).

Chantraine, Pierre, 1933. La formation des noms en grec ancien (Paris: Champion).

Chantraine, Pierre, 1968—1980. Dictionnaire étymologique de la langue grecque (Paris: Klincksieck).

Frisk, Hjalmar, 1960-1972. Griechisches etymologisches Wörterbuch (Heidelberg: Winter).

Furnée, Edzard Johan, 1972. Die wichtigsten konsonantischen Erscheinungen des Vorgriechischen
(Proefschrift) (Leiden: Mouton).

Hamp, Eric, P., 1953. “IE Nouns with Laryngeal Suffix”, Word 9: 135—141.

Kammerhuber, Annelies, 1969. Hethitisch (= Handbuch der Orientalistik, I. Abt. II. Bd. 1 u. 2) (Leiden:
Brill).

Kellens, Jean, 1974. Les Noms -- Racines de l’Avesta (= Beitr. zur Iranistik 7) (Wiesbaden: Reichert).

Kronasser, Heinz, 1966. Etymologie der hethitischen Sprache (Wiesbaden: Harrassowitz).

Leumann, Manu, 1977. Latienische Grammatik 1 (2. Auflage) (München: Beck).

Lejeune, Michel, 1974. Manuel de la langue venete (Heidelberg: Winter).

Mayrhofer, Manfred, 1953. Kurzgefaßtes etymologisches Wörterbuch des Altindischen 1 (Heidelberg:
Winter).

Mayrhofer, Manfred, 1976. Kurzgefaßtes etymologisches Wörterbuch des Altindischen 3 (Heidelberg:
Winter).

Meillet, Antoine, 1934. Le Slave commun (2nd edition) (Paris: Champion).

Nehring, Alfons, 1959. “Idg. *mari, *mori”, in: Festschrift für F. R. Schröder zu seinem 65. Geburtsta-
ge, edited by W. Rasch (Heidelberg: Winter), 122-138.

Pedersen, Holger, 1938. Hittitisch und die anderen indoeuropäischen Sprachen (Köbenhavn: Munks-
gaard).

Pokorny, Julius, 1959. Indogermanisches etymologisches Wörterbuch (Bern/München: Francke).

Reichelt, Hans, 1967. Avestisches Elementarbuch (2. Auflage) (Darmstadt: Wissenschaftliche Buchge-
sellschaft).

Risch, Ernst, 1974. Wortbildung der homerischen Sprache (2. Auflage) (Berlin/New York: de Gruyter).

Schmidt, Johannes, 1889. Die Pluralbildung der indogermanischen Neutra (Weimar: Böhlau).

Schwyzer, Eduard, 1939. Griechische Grammatik 1 (München: Beck).

Szemerényi, Oswald, 1970. “The IE Name of the 'Heart'", in: Donum Balticum to C.S. Stang (Stock-
holm: Almqvist & Wiksell), 515—533.

Thurneysen, Rudolf, 1946. A Grammar of Old Irish (Dublin: Dublin Institute for Advanced Studies).

Tischler, Johann, 1977. Hethitisches etymologisches Glossar (Innsbruck: Institut für vergleichende
Sprachwissenschaft).

Toporov, V.N., 1975. Prusskij jazyk. (Slovar A—D) (Moskva: Nauka).

Trautmann, Reinhold, 1910. Die Altpreußische Sprachdenkmäler (Göttingen: Vandenhoeck & Rup-
recht).

Vendryès, J., 1959. Lexique étymologique de l'irlandais ancien (Dublin: Dublin Institute for Advanced
Studies).

De Vries, Jan, 1971. Nederlands etymologisch woordenboek (Leiden: Brill).

Wackernagel, Jacob — Albert Debrunner, 1954. Altindische Grammatik 2 (Göttingen: Vandenhoeck &
Ruprecht).

Wackernagel, Jacob — Albert Debrunner, 1957. Altindische Grammatik 3 (Göttingen: Vandenhoeck &
Ruprecht).

Four contributions to Sanskrit etymology

Thomas Burrow # (Balliol College, Oxford)

1. kr-, kirdti *to scatter

The usual etymology offered for this Sanskrit verb derives it from an Indo-European root
(s)ker- ‘to spring, jump’, from which are also drived Gk. oxAtpw ‘to skip, dance, frisk' and
a variety of other words in various languages listed by J. Pokorny 1959: 933—934). The
meanings do not correspond, which renders this proposal uncertain. This derivation also
assumes that the final vowel of the Indo-European root involved was -r, which cannot be
taken for granted. It may just as well have been -/. Sanskrit r may be derived from Indo-
European r, or from Indo-European /. On the other hand Sanskrit / does not, as commonly
assumed, freely develop from Indo-European 7. Where / occurs in Sanskrit, whether replac-
ing an r in the Rgvedic language, or without such an alternative, it corresponds, with few
exceptions, to Indo-European /. Although the verb kr-, kirdti only shows r in all stages of
the language, there are a number of words which have been held to be derived from it, or
connected with it, which have 7. If these connections are justified, then the great probabil-
ity is that the Sanskrit root kř- is derived from an Indo-European root ending in A

In the first place there are the adjectives ckula-, ‘confused; filled with, crowded (also
vyakula-, samakula-) and samkula- “crowded together, dense, disordered’. The Petersburg
dictionary derived these words from the root &7-, and the meanings are eminently suitable.
Mayrhofer (1953—1980: s.v. akula-) rejected this etymology in favour of a derivation from
Proto-Munda which probably he would not now be prepared to consider. On the other hand,
under sa mkula- he is prepared to accept it, although ready to assume a development ofr to
l, which, as noted above, is not acceptable. This matter will be taken up later. In the mean-
time we may note the point made by Professor Turner (1966: 45), that there is a remark-
able parallelism between akula-, vyakula-, samkula-, samakula- and akirna-, vyakirna-,
samkirna-, samakirna-, the two sets having the same meaning.

The Petersburg dictionary also derived the adjective kalila- from this root. The word is used
in the sense of ‘dense with, filled with’ at the end of compounds, frequently with the idea
of a confused mass: rathanagasvakalila (vahini), etc. In such cases we could substitute
akula-, akirna-, etc. without changing the meaning, and this confirms the derivation.

In addition to the derivatives already suggested we may further connect with the root kr-
another word showing -/-, namely the tenth class verb kalayati ‘puts to rout, chases, pur-
sues’. This verb is used frequently in the epic language in connection with defeated ene-
mies, and the basic meaning is ‘scatter, disperse’. It provides further evidence that the kr-
has an r corresponding to Indo-European /.

Finally there is the peculiar combination of words kiri- and kiti- both meaning ‘hog’. There
is no doubt about the derivation of the first of these. The root kř- is used of animals digging
up and disturbing earth, and the stem kiri- is a straightforward derivative of this, descriptive
of the activity of the hog. The same stem as an action noun occurs in the compound akhu-
kiri- ‘molehill’. The problem is to find out how the form kiti- can be connected with kiri-.
This can be done if we assume that the Indo-European root originally terminated in -/. In
that case kiti- can be derived from *kilti-, with the suffix -ti added to the /-form of the
root, and this by the operation of Fortunatov’s law would regularly become kifi-.

In considering the etymology of this root we also have to take into account those forms of
it which appear with an initial s-mobile. In such cases the form with initial s- is usually
more original, and the form without the s- is to be regarded as having lost it; pasyati: Lat.
specio; tanyatu-: stanayitnu-, etc. So we must look for comparisons in Indo-European
having initial sk- This of course has been done, but the error has been to assume that the
final consonant of the Indo-European root was -r. As we have seen there is plenty of evi-
dence from Sanskrit itself to show that the original root must have terminated in +.

The forms of this root with initial sk- occur in connection with certain prepositions. These
are as follows:

(1) apa:

The examples are from Panini 6.1.142 where the combination apaskirate describes the
actions of a bull turning up earth with its horns, of a fowl scratching the earth looking for
food, and the action of a dog preparing to lie down. As these actions are looked upon
without regard to any particular object, the verb is conjugated in the middle: apaskirate
vrsabho hrstah, apaskirate kukkuto bhaksarthi, apaskirate sva asravarthi. There is also a
noun apaskara- ‘dung’.

(2) ava:

The only verbal example of this combination is avcaskarire in Sisupälavadha 5.63. This also
depends on Pänini’s rule 6.1.142, and its meaning is the same as that of the combination
with apa (subject mahoksäh). In addition there is a noun avaskara- with the same meaning
as that of apaskara-, ‘dung; a place where dung and other rubbish is deposited’.

(3) upa:

Initial sk- of this root is again attested only in Panini’s grammar, 6.1.141—142. The exam-
ples are upaskaram madraka lunanti, the precise meaning of which is not certain, although
the commentary renders upaskaram by viksipya, and the participle upaskirnam, which is
also obscure, the commentary paraphrasing it as viksepa-.

(4) prati:

This combination is also known only from Panini (6.1.141), and the meaning is equally
obscure. Again the commentary renders pratiskirnam by viksepa-.

(5) vi-:

This combination occurs only in the nominal stem viskira- ‘scatterer’, i.e. a ‘gallinaceous
bird’.

On the basis of the evidence from Sanskrit so far provided we must look for an Indo-Euro-
pean root beginning with sk- and terminating in -/.

In Iranian this root is represented by Av. kärayeiti, where it has been specialised in the
sense of scattering or sowing seed, a meaning also found in Khotanese and elsewhere (Bailey
1979a: 53). The verbal form of Avestan corresponds to that of Skt. kalayati cited above. A
more general meaning is found in Ossetic kalyn ‘pour, scatter’. The fact that Ossetic has --
in this word may be of significance, although it is possible to account for this by develop-
ment within Ossetic itself (Abaev 1958: 570).

A form with initial sk- may possibly be attested in Iranian by the Avestan proper name
skarayat.ra0a- which may have meant ‘scattering the chariots (of the enemy)’.

A verbal root having a comparable meaning, and a phonetic form corresponding to what is
required by the above evidence, is found in Greek: oxdAdw ‘stir up earth, hoe’, oxadew
‘hoeing’, oxadevw ‘stir, poke; (of poultry) scratch’, okaAA&w ‘hoe’, oxador ‘hoeing.
There has been some restriction of the meaning in Greek, so that the use of this verb is
confined to the activity of scattering and disturbing earth, but it is to be noted that this
precise sense is also prominent in Sanskrit, as illustrated in the examples quoted above.

The same root, Indo-European skal-, appears also in Celtic: Irish scailid ‘scatter, disperse’
with a meaning corresponding exactly to that of Skt. kirati. In the dialects of northern
England and Scotland there is a common verb scale having the same meaning. According to
Wright (1855-1930: 570) its meanings are as follows: ‘to spread hay or manure over a
field, to scatter and level molehills, to dress ground by raking and hoeing, to stir, poke,
rake out a fire, to disperse, separate (gathering of people transitive and intransitive); to
spill, upset'. This word appears to have no relatives in other Germanic languages and so
could possibly be a loanword from Celtic, but whether this is so or whether it is a Ger-
manic word preserved only in these dialects of English, there is no doubt that it belongs to
the Indo-European root skal-, as represented in the Sanskrit, Greek and Celtic words listed
above.

The etymologists have usually connected the Greek word with the Indo-European skel- ‘to
split’, which occurs in such forms as Lith. skeliu ‘to split’, Olsl. skilja, Arm. celum ‘id.’, etc.
This is not suitable semantically, since ‘split, cleave’ and ‘scatter, disturb, disperse’ are quite
different meanings. The connection is also unsuitable phonetically, since the forms of the
verb meaning ‘scatter’ unaminously point to an Indo-European root having the vowel -a-,
i.e. skal-. The two groups should therefore be kept separate.

The Celtic forms of the verb meaning ‘to scatter’ in the Britannic section show some com-
plications: Welsh chwalu ‘to scatter’, Cornish skulye, Breton skuld ‘id.’. The initial ch- of
the Welsh word is explained as the result of the metathesis of initial sk- to ks-, of which
there are other examples, but the following -w- is a problem. J. Morris-Jones (1913: 159)
assumed two original forms of this root, sq“el- and sqel- to account for this development in
Welsh, and though he has not been followed in this it seems that some kind of explanation
on these lines should be necessary. Of course he assumed the usual theory that the Indo-
European root in question was skel- (sqel-), which as we have seen is not acceptable, but a
similar variation with an alternative skal- or skWal- would account for this development. In
this case the Cornish and Breton forms of this verb could be explained as having the weak
form of the root skWal- (alternation skWal- : skul-). The same development can also be seen
in Gk. okUAXw ‘to be dishevelled (hair), be torn to pieces of dead bodies (medio-passive);
(transitive active) maltreat, molest, trouble'. This verb has usually been considered to be
connected with okáAAco (Frisk 1970: 742), and its phonetic form agrees with the above
mentioned Celtic forms. This takes back the alternation to an early period.

If such an alternation could occur with the root beginning with s-, it could also occur with
the alternative form without s-; i.e. kwal-/kul-. Of these the former is not attested, but the
weak form kul- can be seen in Skt. akula- and samkula-, and this would account for the un-
expected vowel -u- contrasting with the -i- of kirati.

2. chata ‘mass, lump, assemblage’

This word appears late in Sanskrit being first recorded in Sisupälavadha 1, 47. There is
another meaning, ‘collection of rays, lustre’, but this is secondary arising from such com-
pounds as kiramacchatà ‘mass of rays’ and vidyucchata ‘mass of lightning’. Prakrit has a
corresponding word chada ‘mass, assemblage’, but chada ‘lightning’ given by the Desinama-
mala is again a secondary development arising from the compound vidyucchata. In Bud-
dhist Sanskrit there is a reduplicated form chata-chata or chata-chata used adverbially in
the form chatachataya or chatachataye meaning ‘in one great mass’.

The fact that the word occurs late in Sanskrit does not prevent it from having an Indo-
European etymology. Such words may have been preserved in vernacular Indo-Aryan from
early times, and eventually been adopted into Sanskrit from Prakrit at a comparatively late
date. In looking for an etymology of the word chata we have also now to bear inmind the
fact that intervocalic -f- may by the process of spontaneous retroflexion represent Indo-
European dental -f-, as I have illustrated in detail elsewhere (Burrow 1971: 538-559).
Since initial Sanskrit ch- corresponds to Indo-European sk-, we may posit in this case an
Indo-European skat-, and this will correspond to the scat- of Lat. scateo, of which the
meanings are given as ‘to gush, well, spring or flow forth (of wells, springs, fountains)’, and
also in a more general sense as ‘to abound, be abundant, abound with’. It has been general-
ly taken for granted that of these two meanings the first is original, while the more general
meaning is a secondary development. The Sanskrit comparison would suggest that the
opposite is the case, and this is confirmed by other forms in other Indo-European languages
which have been compared with Lat. scateo (J. Pokorny 1959: 950). Thus Irish scatan
‘herring’ is obviously derived from the fact that these fish occur in very large shoals, and
the same argument applies to some Germanic words meaning ‘spawn’ which are quoted in
this connection. The respective priority of the meanings of the Latin word should therefore
be reversed.

It has been customary to put with Lat. scateo, etc. Lith. skasti (skantu, skatait) ‘to jump,
spring’, the idea being that a meaning ‘to rise up’ is common to Lat. scateo ‘to spring up,
well up’ and Lith. skäsfi ‘to jump’. This connection is impossible when it is seen that the
primary meaning of Lat. scateo is ‘to abound, be abundant’, since the meanings are dif-
ferent. There is a good etymological connection for the Lithuanian verb with Gk. (Hesych.)
toxatduttev ‘jumped’. This is a denominative verb from a *skatamos formed from a root
skat- which is identical with that of the Lithuanian verb. To judge from Lithuanian this
root is different from the previous one, which has palatal k (skat-), in having a velar k
(skat-). The etymological connection of the Greek and Lithuanian words is clear, but they
have nothing to do with Lat. scateo, etc., and they should form an independent entry in an
Indo-European etymological dictionary.

61
3. pitta- n. ‘bile’

No satisfactory etymology for this word has so far been proposed. Since it was thought
that ‘bile’ might be so called on account of its colour, as is sometimes the case (Gk. xoA7
etc.) it has been supposed that there might have been some connection with the adjective
pita- ‘yellow’, but no way has been found to show how these words could be formally
connected. A Dravidian explanation which I had attempted (cf. Burrow, BSOAS 11. 1948:
345; 12. 1949: 385) based on the same semantics connecting it with Ta. pacu/pai, etc.
might have accounted for the irregular phonetic correspondence between pitta- and pita-
which has to be assumed in the case of this as well as of the previous etymology, but it runs
into difficulties equally great when examined in detail. This proposed derivation of pitta-
and pita- was not included in DED 3161.

Any etymology of Skt. pittd- must be based on a definition of its structure. It contains the
accented suffix -td, which is the suffix of the past participle passive. Consequently we must
look for a verbal root in the first part of the word, and this could have been either pit- or
pid-. No such roots are found in Sanskrit, but there is always the possibility that some such
root could have existed in the prehistoric period of the language and though obsolete as a
verb, could have been retained in this derivative. So we must look to find whether any
suitable such root can be found elsewhere in a related language, which would account for
this word. Such a root can be found, and it is in fact pid-.

Greek has the following words showing the root-form pid-: môaë ‘spring, fountain’,
miSaxdets ‘rich in springs’, möaw ‘to swell out, gush forth’ mdvq ‘id.’, and miSvate ‘leak-
ing’. The Greek words have long -;- which contrasts with the short -i- in pid- which is
assumed for the word pittá-. This variation will be dealt with below, but for the present we
are concerned with the meaning of the root. ‘Bile’ is secreted in the liver and discharged
into the gall-bladder, and and it is the result of this process which is denoted by the word
pittd-, derived from a root pid- having an appropriate meaning.

We may also see a derivative of this root in pidaka- ‘boil’. There are two forms of this word
used in Sanskrit, pifaka-, and pidaka-, but it can be seen that the more original form is
pidaka-, by comparison with the Pali form of the word pilaka, where intervocalic -/- corre-
sponds to original -d-. The form pitaka- has arisen by hypersanskritisation. The -d- of
pidaka- is yet another case of spontaneous retroflexion, to which reference has already
been made in connection with the word chata. The original form was *pidaka- from this
same root pid-, and it is so called because it secretes and exudes pus.

The root pid- is an extension of a simpler root pi-, which is in common use in Indo-Aryan
and Iranian: pdyate, etc. ‘to swell, overflow, be exuberant, abound’, used primarily of milk
in the udder, but also in a more general sense. The short -i- of this root is attested by the
forms of the perfect participle active: Vedic pipyusi, Av. pipyusi. The same form of root is
also found in Gk. mus ‘pine’, with the suffix -tu. The pine is so called because it exudes
resin.

There is also a long-vowel extension of this root producing a root pya- (apyayate, etc.),
having the same meaning as the simpler form of the root. The weak grade of this extended
form of the root is pi-, as occurring in the past participle äpina- (with an alternative apya-
na-). This form with long -i- is also found in Lat. pituita ‘discharge from the nose’; also in
Lat. pinus ‘pine’, and in Skt. pitu-daru- ‘a kind of tree (deodar or khadira?)’. It is also this
form of the root which has received the extension -d in Gk. riag, etc.

The root-forms pi-, pyä- and pid- and their derivatives form a mutually consistent set with
common basic meaning. In particular the words for ‘pine’ are straightforwardly derived
from this root, and there is no reason to suppose that they are not Indo-European in origin
(Benveniste, BSL 51.1954: 31) or that they are an ‘Indo-Mediterranean group’. In the ety-
mological dictionaries the above words have been associated with other word-groups which
should be separated from them. Among these are the words for ‘fat’, Skt. pivas-, etc. The
meaning of this series is quite different inasmuch as fat is a solid deposit and is not dis-
charged. The kind of verbal root that would be suitable for a word meaning ‘fat’ is illus-
trated by the word vapa. I have dealt with this word elsewhere (Burrow, BSOAS 45.
1982: 188), and shown that the meaning is not ‘omentum’ as usually given in the diction-
aries, but the internal deposit of fat inside an animal and that it is derived from the root
vap- which means ‘to cast down, deposit’ particularly where an accumulation results from
this action (as in vapra-). The word in this sense is therefore not different in derivation
from vapd used in the sense of heap in valm ika-vapá 'anthill. By contrast the root pi- and
its extensions are quite unsuitable to form the basis of a word meaning ‘deposited fat’. The
root pi- of Sanskrit pivas-, etc. does not, in fact, appear to be a verbal root.

Another word which should be kept separate is pifu- ‘food’. This has been attached to the
above set mainly as a result of a wrong statement of meaning (‘Saft, Trank’: Pokorny
1959: 793--794). Its only meaning is ‘food’, and it refers primarily to solid food. The word
pitu- is to be derived from the Indo-European root pa- ‘to feed’, and it contains the root in
its zero grade followed by the incremental vowel -i- and the suffix -tu (cf. further Burrow
1979: 44).

The adjective pita- ‘yellow’ is still without certain etymology, but from what has been said
it is clear that it can have no connection with pittá- ‘bile’. H.W. Bailey compares Khotanese
pe ‘colour name, green(?)' (Bailey 1979a: 248), which is possible, but since the meaning is
conjectural, somewhat uncertain. Another possible connection would be with Gk. (Hesych.)
TULAUG * TMGPAAEVKOS.

4. bidala- ‘cat’

Of the two common words for ‘cat’ in Sanskrit the word marjara- is clearly derived from
the root mrj-/marj- ‘to cleanse, wipe’, but the derivation of bidala- has remained uncertain.
According to M. Mayrhofer (1963: 429) it is ‘‘nicht sicher erklart; wohl ein Fremdwort”.
The normal form of this word which is also the more original form, is bidala-, correspond-
ing to which there are Pa. bilala- and Pkt. bidala-. Pali has also a form bilära- showing the
familiar alternation of r with /. There are late lexical forms bilala- and birala-. Of these the
first -/- in bilala- is the -/- which in Prakrit corresponds to the -/- of Pali, and therefore to the
-d- of Sanskrit. In birala- there is a metathesis of the consonants of Pali bilara- and a corre-
sponding Pkt. *bilara-. This is the form most commonly represented in the modern lan-
guages (Nep. biralo, etc., Turner 1966 no. 9237). There is also a hypocoristic billa, billi
‘cat’ of which the latter form has been adopted into Dravidian (DED. no. 3438).

An attempt was made by me to compare this word with the Dravidian words for ‘cat’ (Ta.
veruku, etc. DED 4520). There were two faults with this comparison which make it unac-
ceptable. In the first place it assumed that the form birala- was the original form, but it is
quite clear that this is a secondary development in Indo-Aryan. In the second place only

the first syllable of the Dravidian word was compared, but in all Dravidian languages with-
out exception this syllable ver- is followed by a -k-, which is therefore an integral part of
the Dravidian word. Consequently any loanword from Dravidian would be expected to
show this -k-, as has actually happened in Nahali, which has borrowed this Dravidian word
in the form berko/berku.

Before going into the question of the etymology of Skt. bidala- it is necessary to say a few
words about the historical phonology of Sanskrit b. It has recently become increasingly
obvious that Indo-European w (u) which normally appears in Sanskrit as v, is alternatively
in quite a number of instances represented by b. In some cases where this happens there is
an alternative form with v, as is the case with this word which is also written vidala-. In
other cases only the b-forms are attested. Examples of this b < v are as follows:

baskaya- epithet of growing calf « *baksaya- « *vaksaya-; cf. (Burrow, JRAS, 1969: 114).

bata- ‘bad’; Khotanese bata- ‘small; reverse to good’, Zor. P. wt vat ‘bad’, etc., H.W. Bailey
(1979a: 95).

barsva- ‘gums of the teeth’ < *wolswo- : cf. Gk. oöXos (Studies... L.R. Palmer, Innsbruck,
1976: 40).

bali- ‘offering; tax’: Khot. gvar- ‘to distribute’, Ossetic D iuarun ‘distribute, divide’, etc.,
Bailey (1979a: 95).

bala- ‘ignorant, foolish; child’: cf. OSlav. volu ‘ox’, cf. (BSOAS 38. 1975: 76).

baspa- (also vaspa-) ‘tears; steam, vapour’ « "vapsa-; Lat. vapor/vapos; cf. (RAS 1969:
114).

The Sanskrit word. bala- n. 'strength, force’ is instructive in this respect. H. Grassmann
derived it from *vala-, and compared Lat. valor. This was unsatisfactory to the Junggram-
matiker, since it went counter to their doctrines that sound rules had no exceptions, and it
was clear that the normal representation in Sanskrit of Indo-European w fu) was v. As a
result an alternative etymology was proposed, connecting it with Gk. BeATiwv > Lat. debilis,
and OSlav. boliju. Considering this to be unsatisfactory for various reasons, an alternative
derivation from Dravidian was proposed by me (Burrow, 7PS, 1946: 19) connecting it with
Ta. valam, etc. (DED 4317). The main objection to this is that a word so central to the
Vedic vocabulary is unlikely to have been a loan from a non-Aryan source. Furthermore,
the existence of comparative and superlative forms from this root (balıyan balistha-) is
evidence that an Aryan root is involved. These forms would not have been derived from a
neuter noun which was a loanword. For these reasons Skt. bala- is not being included as a
loanword from Dravidian in the revised edition of the Dravidian Etymological Dictionary.
After all this H.W. Bailey has returned to the etymology of H. Grassmann, and he now pro-
poses to connect Skt. bala- with the Indo-European root ual- ‘to be strong’ (Bailey, BSOAS
42.). In view of the frequency with which Indo-European w /u) is now known to be rep-
resented by b, there is no doubt that he is correct in doing this, and that the earliest ety-
mology of bala- is the right one.

We have also to consider the nature and origin of the intervocalic -d- in bidala-. It has al-
ready been pointed out that a dental in such a position is frequently converted to a retro-
flex by a spontaneous process. Furthermore when such a retroflex cannot be explained by
combinatory change (as in nida-, etc.), then it must be a case of spontaneous change in

words which are originally Indo-Aryan. Such was the case with the words chata and pidaka-
treated above, and this is the case here. Taking this into account, and also the fact that the
initial b- is from v- (which also occurs), we get back to an original form *vidala- This can
be analysed as vi-dala-, of which the first member is the word vi- ‘bird’, and it means the
animal which attacks and tears to pieces (small) birds, which is one of the most character-
istic habits of the cat. An alternative analysis, taking vi- to be the usual verbal prefix vi
would not be specific enough to provide such a name.

The verbal root involved is dal-, dalati ‘burst, split’ caus. dalayati and dalayati. This has
been considered to be a variant of dr-, drnati from Indo-European der- (Mayrhofer 1963:
24). In this case the form with -r- (Pa. bilara-) would be the more original. On the other
hand there is also evidence for an Indo-European root del- with the required meaning
(Pokorny 1959: 194), and since the occurrence of -/- in Sanskrit usually indicates Indo-
European -/- this may be the root which is represented in Skt. dalati, dalayati, and conse-
quently in bidala-.

References

Abaev, V.I., 1958. Istoriko-etimologiteskij slovar’ osetinskogo jazyka 1 (Moskva: Nauka).

Bailey, H.W., 1979. Dictionary of Khotan Saka (London: Cambridge University Press).

Burrow, Thomas, 1971. Spontaneous cerebrals in Sanskrit", BSOAS 34: 538—559.

Burrow, Thomas, 1979. The Problem of Shwa in Sanskrit (Oxford: Clarendon Press).

Burrow, Thomas/Murray Emeneau, 1961. A Dravidian Etymological Dictionary (Oxford: Clarendon
Press).

Bóhtlingk, Otto/Rudolf Roth, 1855—1857. Sanskrit-Wórterbuch (St. Petersburg: Kais. Akad. der Wis-
senschaften).

Frisk, Hjalmar, 1970. Griechisches etymologisches Wörterbuch 2 (Heidelberg: Winter).

Mayrhofer, Manfred, 1953—1980. Kurzgefaßtes etymologisches Wörterbuch des Altindischen (Heidel-
berg: Winter).

Morris-Jones, J., 1913. A Welsh Grammar, Historical and Comparative (Oxford).

Pokorny, Julius, 1959. Indogermanisches etymologisches Wörterbuch (Bern/München: Francke).

Turner, Ralph, 1966. A Comparative Dictionary of the Indo-Aryan Languages (London: Oxford Uni-
versity Press).

Wright, Joseph, 1855—1930. English Dialect Dictionary (London: Frowde).

On Sanskrit bhunakti ‘aids, serves, protects’

George Cardona (University of Pennsylvania)

1. Panini (Astadhyayi 1.3.66: bhujo’navane) provides that bhuj takes atmanepada affixes
if the verb is used in a meaning other than ‘aid, protect’ (anavane); that is, in the Sanskrit
of Panini’s time one used middle forms bhunkté etc. meaning ‘enjoy, use, consume’, but
active bhundkti etc. ‘aid, serve, protect’. Now, Panini derives bhunkté and bhundkti from a
single verb listed in his verb catalog (his dhatupatha) but bhujati ‘bends’ from another
verb, for formal reasons: bhunkté, bhundkti are formed with the same affix, and bhujati
has a different affix. Panini does not make any other formal connection between bhuñkté
and bhundkti. In particular, he does not treat the latter as somehow a causative with
respect to the former; on the contrary, only the formal causative bhoji (bhojayati etc.) is
considered to bear this relation with bhunkté. This state of affairs reflects a historical
development which I shall consider here briefly.

2. In early Vedic, as is well known, bhunkté etc. are construed regularly with instrumental
forms referring to whatever serves to give one enjoyment, benefit or nourishment.! The
sources of enjoyment can be as general as ones desires (kamaih, e.g. Maitrayani-samhita
[MS] 4.2.1 [21.19]: ... sérvaih kämair bhunñkte) or as vital as life and breath (e.g., MS
3.3.1 [33.1]: ... Zyusa ca pranéna ca bhufijáte). Among the things Vedic poets seek for
their benefit is the help of the gods; e.g., Rgveda (RV) 8.67.16: sásvad dhí vah . . . dditya
utibhir vayam /purá nünám bubhujmáhe *Adityas, ever have we enjoyed your help, before
and now.’ With the help of gods, one has, for instance, cows to subsist on; thus, Indra is
asked to return cows that have gone off, RV 10.19.6cd: punar na indra gà dehi /fvàbhir
bhunajamahai ‘Indra, give us back our cows, we would enjoy them alive.' Living cattle were
of course used as draught animals, to give milk, and to give more cattle. They could also
be eaten; e.g., Kausitaki-brahmana (KB) 11.2.29—32: tad yatha ha va asmirñl loke manusyah
pasun asnanti yathaibir bhufjate evam amusmiml loke pasavo manusyan asnanti evam ebhir
bhufjate ‘As in this world men eat animals, as they make use of them, so in yonder world
animals eat men, make use of them.” As expected, bhuj is also used with general terms
referring to food one consumes (e.g., ánnena in Taittirlya-samhita (TS) 6.2.5.4.: . . . dnnena
bhufijate).

3. It has equally been known for a long time (see, e.g., PW V.303) that active bhundkti etc.
can be construed with accusative forms referring to what is served by another.

3.1. TS 2.6.2.3—4? concerns an invitational verse and an offertory verse (puronuvakyd,
yajya) which together have seven verse sections, so that they constitute a saptapada
Sakvari, a potent (sdkvari) strophe, by whose power (ydsyai viryéna) certain things are
done: One defeats enemies already born and wards off in advance enemies yet to be born,
one creates light for oneself in both worlds, an ox helps (bhunakti) with its forequarters, a
cow with its hindquarters.° The power of the sakvari strophe is apportioned between the
puronuvakya and the yajya verses. The name of the deity addressed, Agni, occurs in the
first part of the former (agnih in pada a), in the last part of the latter (agne in pada d), so
that the puronuvakya and the yajya are said to have their characteristic marks — the name
of the deity — respectively in front and in back. Thus, with the power of the sakvari
strophe that resides in the puronuvakya one defeats just those enemies that are born
beforehand, with the power in the yajya those enemies that are to be born later; similarly,
by the power in these two, one sets down for oneself luster in this world and in the world
yonder, respectively, and both worlds are full of light for him who knows thus. Moreover,
it is because of this division of power that an ox helps its master with its forequarters, a
cow with its rearquarters, and both of these serve him who knows thus: yá evám véda
bhurktd enam etau.®

3.2. The cow and ox are also said to be of greatest use in a passage from the Satapatha-
brahmana? which gives an explanation of why the Yajamana is not to partake of these
animals when the Adhvaryu has him go to the ritual shed in the course of the Agnistoma.
The cow and ox maintain all this that there is. The gods said that, since the cow and ox
maintain all there is, they should put in them the vital strength of other animals. They did
so. Therefore, the cow and ox most help one: tásmad dhenus caivanadvams ca bhuyistham
bhunktah.? If the Yajamäna partook of these creatures, in which the essential strength of
all animals was put, then, he would be eating all there is, including the fetus of his wife, if
she were pregnant, so that people would say of him that he had killed this fetus, committing
a sinful act, thus acquiring a sinful reputation.

3.3. Other brahmana passages speak of cattle (pasdvah) serving one. A group of related
texts concerns an offering cake that, in the course of the Vaisvadevaparva rite in the Catur-
masya, is taken from its kapala and put in a vessel into which ghee is poured. The cake is
equated with the Yajamana, the ghee with cattle. The Taittiriyabrahmana? says that by
pouring ghee over the cake in the vessel, the Adhvaryu makes the Yajamana rich in cattle.
Now, if the Adhvaryu poured only a little ghee, only few cattle would come to the Yaja-
mana (upatistheran ‘would approach as subjects"), so that they would not really help him
(abhufjantah ‘not serving’). On the other hand, if the Adhvaryu pours a lot of ghee, many
cattle come to the Yajamana helping him (bhufjdntah). Hence, the Adhvaryu should pour
a lot of ghee, but leave the top of the cake exposed: thereby, many cattle come to the
Yajamana helping him. Other texts bring up a possible question about why the cake should
be left with its top exposed. One could worry that, if the Adhavaryu completely filled up
the pot with the cake in it, thus covering the cake, he would make the Yajamana lower
than the cattle, so that they would not serve him when they came to him. Therefore, one
should leave the top of the cake uncovered. This lets the Yajamana stand above the cattle,
which thereby serve him. The texts conclude, however, by saying this is not something to
worry about, so that the pot should be filled; for cattle never do not serve their masters:
nd hi pasavo nd bhufijánti.

3.4. Of course, cows serve their masters by giving milk and bearing calves, oxen by working
as draught animals (see notes 2, 5, 9). Further in Indian imagery, as is well known, cows
are also grant-wishing animals (kamaduhah, kamadughah, kamadhenavah). Now, the Yaja-
mana invokes the fire deity, Agni, asking that the bricks piled to make the altar be cows for
him in this world and the one beyond, thus helping him in both worlds.!! As the Maitraya-
ni-samhita puts it, these bricks are cows that Prajapati used to milk whatever he desired;
the Yajamana similarly uses them to milk whatever he desires.!?

3.5. Another well known ritual image is that of the sacrificial fire (agni) as transporter
(vahni) of gods and offerings, and this is linked to the Yajamana's giving a draught ox as a
ritual fee after the ritual fire has been set up. The sacrifice is the yoke-piece or burden that
Agni bears as transporter, so that the fire, once it has been set up, aids the Yajamana
(bhunakty enam agnir ühitah) by serving as transporter. P?

A ritual implement too can be said to help in a rite. Thus, the ladle called darvi is asked to
fly to the gods when it is full of oblations, then to return, bringing with it the fruits of the
offerings: It is to bring to the sacrificers nourishment and strength, thus aiding all the
sacrifices. 4

3.6. On a more mundane level, bhundkti is used in speaking of an offspring that cares for a
parent. The Taittiriya-brahmana!5 relates a myth that Prajapati gave rise to creatures
through the Vaisvadeva rite and that these offspring did not procreate. Agni then desired
to make the offspring procreate. He caused Prajapati pain, by reminding him that his crea-
tures did not procreate, and he himself also suffered pain, since he desired but did not have
offspring. Thus it is that both he whom an offspring cares for (vám ca praja bhundkti) and
he whom no offspring cares for both suffer pain.!é

Those who care for one can also be members of a full household. In the Sankhayanaranyaka
(SA) and the Kausitaki-brahmanopanisad (KBU)!? a comparison is drawn between how the
knowing self is related to its subservient selves — speech, sight, hearing, mind — and how
the head of a household is related to members of his household. The selves subservient to
the knowing self follow him as members of a household follow their head; and as the head
of a household enjoys the use of its members (yatha sresthi svair bhunkte) and these mem-
bers serve the head (yatha va svah sresthinam bhujfijanti), so does the knowing self make
use of the subservient selves (evam evaisa prajfatmaitair atmabhir bhunkte) and these serve
this self (evam evaita atmana enam atmanam bhufijanti).'®

Just how the subservient selves might be considered to assist the knowing self is brought
out in a passage from the Kausitaki-brahmana!? that concerns invitational and offertory
verses which, instead of being recited according to their order in the Rgveda, are mixed:
RV 4.10.1 is paired with 4.10.3 and RV 4.10.2 is paired with 4.10.4. A parallel is drawn
with the breaths, which are said to serve the physical self as though mixed (vyatisikta iva
va ime prana atmanam bhufijantiti), that is, related to each other, with each one doing its
proper job.

4. The relation between the middle bhunkté etc. and the active bhunakti etc. in usages
such as those given is fairly obvious and can be shown schematically by the following:

l 2
(a) Nnom. Ninstr.

(b) N2 n. N! _ bhuj ot, “Y helps, benefits X.’

bhuj ni d. ‘X helps, benefits himself through Y

This has been recognized not only by modern scholars?! but also by earlier Indian com-
mentators. For example, commenting on KB 11.2.30, Udaya paraphrases ebhir bhufijate
with ebhih pasubhih . . . atmanam palayanti ‘care for themselves through these animals . . ?
(see section 2 with note 2). Commenting on part of a verse cited in Yaska's Nirukta,??
Durga paraphrases na bhunakti with na palayati *does not aid' and explains that this means
traditional learning handed down to bad students does not bring these students together
with the fruit of this learning.”

4.1. There are related sentences with middle and active forms of vrdh which are comparable
to examples of constructions (a) and (b). For example, RV 6.37.5b: indro girbhir vardhatam
... ‘May Indra .. . grow strong through the hymns’ has a middle form (3sg. imper. vardha-
tam), and RV 7.99.7c: várdhantu tva... giro me ‘(Visnu,) let my hymns ... make you
grow strong’ has an active form (3pl. imper. vardhantu), the first construed with the in-
strumental girbhih, referring to hymns through which a god is to grow strong, the second
construed with a nominative (girah) denoting hymns that make a god grow strong and an
accusative (fva) referring to that god. In addition, however, the formal causative vardhi can
be used as equivalent to active forms of vrdh; e.g., RV 5.11.5cd: tvam girah . . . vardháyan-
ti... ‘(Agni,) hymns make you grow’ has girah . . . vardhdyanti with the accusative tvam,
just as RV 7.99.7c has vardhantu ... girah with the accusative fva. The causative vardhi
can also be used in another way, asin RV 1.91.1 lab: séma girbhis tva vayám vardháyamo
... ‘Soma, we make you grow strong with hymns.’ This involves not merely a god and
hymns but also third participants, those who use the hymns to make the gods grow strong.
From its earliest attestation, in the Atharvaveda (AV), the causative bhoji functions in this
system of three participants; e.g., AV 19.50.6c: ... etád asman bhojaya (Night) let us
enjoy that’,”* Latyayana-srauta-sutra 3.6.8: sarpirmadhubhyam rtvijo bhojayet ... ‘He
should feed butter and honey to the officiants.’ In contradistinction to vardhi, then, bhoji
is not used as an equivalent to an active form of the base verb. It is instead linked to the
middle bhunkté etc. separately .?°

4.2. The pairing of bhojdyati etc. with bhunkté etc. is comparable to the pairing of tarpdy-
ati ‘makes satisfied’ with trpyati ‘is satisfied’, from another verb in the semantic sphere of
enjoyment. For example, TS 2.5.11.3--4: ... trpyati prajáya pasubhir ya evam véda ‘He
who knows thus gains satisfaction with offspring (prajdya) and cattle (pasubhih)’, TS
2.5.4.3: ... trptd evaindém indrah prajdya pasubhis tarpayati ‘Satisfied, Indra gives him
satisfaction with offspring and cattle.” As can be seen, both #pyati and farpäyati are con-
strued with instrumental forms such as prajáya, pasübhih.?9 Similarly, bhunkte and bhojá-
yati can both be construed with instrumentals such as kamaih (see section 2) and sarpir-
madhubhyam (see section 4.1.).

4.3. On the other hand, from quite early on, bhoji could be used with both an accusative
denoting someone allowed to enjoy something and another accusative denoting the object
enjoyed; for example, in AV 19.506c (see section 4.1), bhojaya is construed with asman
‘us’ and etdd ‘that’. In this way, bhoji was treated in the same way as causatives from other
verbs in the semantic field of enjoyment, namely verbs of consuming, such as pa ‘drink’.
For example, in Katyayana-srauta-sutra 23.4.24: mrtam brahmanan bhojayet ‘He should
have some Brahmanas eat the dead animal’, bhojayet is construed with the accusatives
brahmanan and mrtam, as in RV 3.57.5cd: .. . thd visvam... ydjatran a sadaya paydya ca

mádhuni ‘(Agni,) have all the worshipful ones sit here and have them drink the sweet
drinks’, paydya is construed with ydjatran and mádhuni. In addition, the base verb bhuj
(bhunkté etc.) came to be used, as were other verbs of consuming, with accusatives denoting
objects consumed; e.g., Sänkhäyana-grhya-sutra 3.1.17; ipsitam annam tad ahar bhuiijita
‘That day, (the student) may eat what food he wishes.’ In contrast to TS 6.2.5.4, where
bhufjate is construed with the instrumental dnnena (see section 2), here bhufijita is con-
strued with the accusative annam?? Moreover, bhuj in the sense of ‘enjoy, use’ also came
to be used in this way. For example, in Raghuvamsa 15.1: krtasitäparityagah sa ratnaka-
ramekhalam | bubhuje prthivipalah prthivim eva kevalam ‘After he had given up Sita, he
(Rama) as king enjoyed only the earth engirdled by the oceans', bubhuje is used with
prthivim. Rama now enjoyed only the earth he ruled as king, not also Sita, whom he had
given up. Similarly, in Manu-smrti 8.150: yah svaminananujfiatam adhim bhunkte 'vicaksa-
nah [ tenardhavrddhir moktavya tasya bhogasya niskrtih ‘A fool who uses a deposit without
the owner's permission is to give up half the interest as compensation for such use’, bhunkte
‘uses’ is construed with adhim.

4.4. The oldest state of affairs for bhuj in Indo-Aryan, such that bhunkte and bhunákti
were used in related constructions as shown in (2), (b) of section 4, thus was superseded by
a new system, in which bhuñkté and bhunäkti were no longer so directly related syntact-
ically. The seeds for this change were present in the earliest system. To begin with, the
pairing of bhunkté with bhundkti in constructions (a) and (b) was not matched by many
similar pairings in any productive system. Even várdhate : várdhati, moreover was an im-
perfect match, since the derived causative vardhi (vardháyati) could be used an an equi-
valent to várdhati but bhoji (bhojayati) was not equivalent to bhundkti (see action 4.1.).
Further, bhunkté not only remained comparable in its syntactic behavior with other verbs
of enjoyment, such as trpyati (see section 4.2. and note 26), it also came to be used with
accusatives denoting objects of enjoyment and consuming (see section 4.3.). The result is
the state of affairs described by Panini.”*

Notes

1 See already the Petersburg lexicon (Bôhtlingk — Roth 1855-1875 [PW]) V.301. Note, however, that
this and other lexicographic sources give a picture that is not completely faithful to the philological
facts. PW V.301 lists first the active bhunakti, then the middle bhunkte, both for the primary mean-
ings given: ‘(1) geniessen, Etwas zu geniessen haben, sowohl zu Nutzen haben, mit Vortheil besitzen,
als vom Genuss von Speisen.’ Following suit, Monier-Williams (1899: 759) gives bhunákti and
bhunkte, in that order, and glosses, ‘to enjoy, use, possess, (esp.) enjoy a meal ...', although he
notes parenthetically that the verb mostly takes middle suffixes in these meanings. More recently,
Mayrhofer (1963: 507) lists only the active bhundkti as the main entry glossed 'geniesst, benützt,
verzehrt/enjoys, uses, consumes', although he goes on to cite middle forms of the verb. See also
notes 6, 8, 10, 14, 16, 18.

? In his commentary on KB 11.2.2.30 (edited by E.R. Sreekrishna Sarma 1976), Udaya (1976: 271)

glosses bhuñjate with atmanam palayanti (see section 4) and explains that men care for themselves

through these animals, the cattle they have, by having them serve as draught animals, milking them,
and using them to bear other cattle: yatha va ebhih pasubhih vahadohaprasavadvaratmanam palayan-
ti.

... kesinam ha darbhyém kest satyakamir uvaca saptápadam te sékvarim svó yajfié prayoktase yásyai

viryéna prá jatan bhratrvyan nudáte práti janigsyámanan yásyai viryénobháyor lokáyor jyótir dhatté

ydsyai viryéna purvardhénanadvan bhundkti jaghanardhéna dhenur (til purástallaksma puronuvakya

bhavati jütan evá bhratrvyan pra nudate/updristallaksma yajya Janisyámanan evá práti nudate/
purästallaksma puronuvakya bhavati asminn evá loké jyótir dhatte [upáristallaksmā yājya amúşminn
eva loké Jyötir dhatte /jy6tismantav asma, imat lokaù bhavato yd evám véda/ {purästallaksma puronu-
vak ya bhavati tásmat pürvardhénanadván bhunakti/ upáristallaksma yajya tésmaj jaghanardhéna
dhenüh/ yá evám véda bhunktá enam etai .

Respectively agni mürdha divéh kaküt / pétih prthivya aydm lapam rétamsi jinvati and bhüvo yajfid-
sya retasas ca neta yätra niyüdbhis säcase sivabhih / divi mürdhanam dadhise suvarsam jihvam agne
cakrse havyavaham, which originally occur as RV 8.44.16, 10.8.6 and are repeated in Taittiriya-
brahmana (TB) 3.5.7.1 and elsewhere.

That is, an ox serves its master by pulling the plow or cart with its shoulders, a cow by giving milk
with its udders. Sayana is, I think, perfectly right when, in his comments on TS 2.2.6.3 (Anandasra-
ma Sanskrit Series 42.2 (Pune 1966:1389)) he says purvardhéna refers to the front part of an ox, its
shoulders, jaghanardhena to the rear part of a cow, its udders, and when he glosses bhunakti with
palayati ‘helps’: ... tadiyena viryena balivardo langalasakatavahina skandharüpena pürvabhagena
svaminam palayati /dhenus ca kgirapradenodhasaparabhagena svaminam palayati. See also note 6.

A. Berriedale Keith (1914: 207) translates bhundkti ‘feeds’ and bhunktäh ‘enjoy’: ‘... by the first
half of whose strength the ox feeds, by the second half the cow ... therefore the ox feeds with its
first half ... therefore the cow feeds with the second half. Him who knows thus these two enjoy.’
This makes little sense in the context of the passage. In addition, ydsyai viryéna purvardhéna .
jaghanardhéna cannot justifiably be translated ‘by the first half ... by the second half of whose
strength.’

SB 3.1.2.21: athainam salam prápadayati /sá dhenvaí canadühas cá nasniyat / dhenvanadühau ' va idam
sérvam bibhrtah [té deva abruvan dhenvanadühau va idam sárvam bibhrtah hánta yád anyésam váya-
sam viryam tad dhenvanaduhayor dédhaméti / sa yad anyesam väyasam viryam asit tád dhenvanadu-
háyor adadhuh tasmad dhenüs calvanadvams ca bhüyistham bhunktah tád dhaitat sarvasyam iva yó
dhenvanaduhdyor agniyat/dntagatir iva tam hadbhutam abhijanitah jayayai gárbham níravadhid iti
papám akad íti papi kirthih [täsmad dhenvanaduhäyor nasniyat.

It is neither contextually nor formally justifiable to translate bhunktah ‘eat’ (Eggeling XXVI.11:
*... and therefore the cow and ox eat most.’) or ‘fressen’ (PW V.302: ‘fressen am meisten’).

TB 1.6.3.4--5: ... ydjamano va ekakapalah pasäva ajvam/ydd ékakapala ajyam andyati yájamanam
eva pasübhih maradna vati] yéd dipam ānáyed álpa enam pasávó ‘bhuñjanta üpatistheran/yéd bahv
andyed baháva enam pasävo bhunjänta üpatistheran/ bahv aniyavihprstham kuryat/bahäva evainam
pasávo bhufijánta üpatisthante. In his comments on TS 1.8.2. (Vaidika SamSodhana Mandala edition
2:319 (Pune 1972: 319)) Sayana cites this passage, glosses bhufijéntah with palayantah ‘helping’,
and specifies that the cattle help by giving milk and similar acts (bhufijantah ksiradanadina palayan-
tah).

MS 1.10.7 (147.2—6): yajamano va ékakapalah pasévo ghrtam/tad abhipüryah /pasübhir_ evainam
sämardhayati/yad abhipürayed ádharam enam pasubhyah kuryat/ abhufjanta enam pasáva upatist-
heyuh [avíhprsthah karyàh [pasübhya evaínam úttaram akah /bhuñjánta enam pasáva üpatisthante/
tán ná surksyam Jabhipürya eva/nd hi pasävo nd bhufjänti. Similarly, Käthaka-samhitä (KS) 36.1
(69.6—10). From such passages it is clear that bhufijantah of T andya-maha- brahmana 25.1.13:
tasmad arvafico bhuñjantah pasava upatisthante cannot justifiably be said to mean ‘eating’, as PW
V.302 would have it (line 6); the cattle approach with heads lowered (arvaficah), serving their masters
by giving milk and such, as Sayana correctly says (... bhufijantah payaadina palayanto gavadipasa-
vah arvafico dhomukhah sayam svamigrham upatisthante).

E.g., Väjasaneyi-samhita 17.2: ima me agna istaka dhenávah santu ... eta me agna istaka santv
amutramismimi loké; $B 9.1.2.17: . tathò hainam eta ubháyor lokáyor bhufijanti asmím$ camus-
mims ca; Sayana on the latter: enam yajamanam bhufijanti pala yanti .

MS 3.3.4 (36.3- 4): dhenávo vai namaita istakah / etabhir vat prajépatir yád yad ákamayata tát tad
dduha | yád yad evattabhir yájamanah kamayate tat tad duhé.

KS 8.5 (89.5), Kapisthala-katha-samhita 7.1 (84.5--7): vahi deyah/vahino vai pasavo bhufjanti/
yajito 'gner vahah /bhunakty enam agnir ahitah. 'The sentence bhunakty enam agnir ahitah is given
in the Kasika-vrtti to Panini 1.3.66 (see section 1) as an example of bhunakti meaning ‘aid, protect’.
Note that I have adopted the reading vahah (. . . vaho bhunakty ... in a samhita reading). Both von
Schroeder in his edition of the Kathaka (1900: 89 with note 2) and Raghu Vira in his edition of the

Kapisthala-katha-samhita (1968: 84 with note 4) adopt yajfio gner vahi, although the former men-
tions that vahoh occurs in his manuscript D and the latter mentions that veo is the reading of his
only manuscript. I have adopted vahah (vaho) both because this has manuscript support and because,
at least to me, it is the reading that makes sense. As is known, vaha denotes an ox’s shoulders (cf.
Amarakosa 2.9.63: skandhapradesas tu vahah), on which the ‚animal carries the yoke (cf. Panini
3.3.119). AV 4.11.8ab: madhyam etad anaduho yatraisa vaha ahitah ‘This is the middle of the ox,
where this vaha has been put’ refers to a vaha put on the middle of an ox, that is, on the shoulder
region. Säyana interprets vahah as meaning ‘load’ (bharah), and others have taken it to designate the
piece of the yoke that goes on the shoulders (Wackernagel-Debrunner 1954: 68: ‘Schulterstück des
Jochs’). In either case, véha refers to something an ox carries. And Agni the transporter (vahni) is
equated with a draught ox in a passage related to the two Yajurveda texts under discussion. TB
1.1.6.10—11 says the Yajamana should give the Adhvaryu an ox (anadvaham adhvaryäve), then
equates both the ox and the Adhvaryu with the fire as transport (véhnir va anadvan/véhnir ad-
hvaryüh): The ox and Agni are both transporters, the first for the cart, the second for ritual sub-
stances (Sayana: anadvan sakatam vahati /agnis ca dravyam vahati);, the Adhvaryu too is a vahni, in
that he "carries out” rites (Sayana: adhvaryur apy anusthanam vahati). Through the ox that is a
draught animal the Yajamana gets a transport for the ritual (véhninaivé vahni yajñasyavarunddhe).
It is understandable that a Brahmana text speaks of the sacrifice (yajfíah) as the vaha (vahah) of
Agni, but it is not immediately understandable that such a text should speak of the sacrifice as the
draught animal (vahi) of Agni. Hence my choice. To be sure, one could adopt the reading ya/fio "
gner vahi and take this to mean "The sacrifice has a bullock's shoulder in respect of Agni’, that is, it
has Agni's shoulders to bear it, but this appears tortured, especially since vahin regularly denotes a
draught animal, not just anything that has an ox’s shoulder.
AV 3.10.7c-f: purna darve pärä pata supurna pünar à pata / sárvan yajfian tsambhuñjatisam ürjam na
a bhara, Whitney’s translation (1905: 1.101) ‘jointly enjoying’ certainly lacks any support in the con-
text of a single ladle.
TB, 1.6.2.1.: vaisvadevéna prajapatih praja ı asrjata/ tah srsta na prajayanta / sÔ entr akamayata ahäm
imah prájanayeyam iti/sé prajapataye sócam adadhat | sò socat prajam icchámanah /tasmad yam ca
praja qd bhundkti yam ca né tav ubhat socatah.
While explaining this passage, in his commentary on TS 1.8.2 (Vaidika Samsodhana Mandala edition,
11.317), Sayana again paraphrases bhundkti with palayati. In RV 8.1.6ab: vésyam indrasi me pitùr
utá bhratur ábhufijatah the poet says to Indra, 'You are worth more to me than my father or a
brother that does not help.' The allusion is most probably to a brother who does not contribute
properly to the maintenance of the joint household; compare the reference to a wealthy man who
does not help others (RV 1.120.12b: ébhufijatas ca revatah).
SA 6.20, KBU 4.20: tam etam atmanam eta ütmano nvavasyanti yatha sresthinam svah/tad yatha
sresthi svair bhunkte yatha va svah sresthinam bhufijanti evam evaisa prajfiatmaitair atmabhir bhunk-
te evam evaita atmana etam atmanam bhufijanti.
Sankarananda (1968: 124) has the reading bhufijate and explains the simile as follows: The members
of the household eat the head of the household by using his substance (dravyadvara ; cf. English 'eat
out of house and home’). Cowell adopts the reading bhufjanti but translates (1968: 171) ‘and as the
household feed on the householder’, which is not acceptable; better, Renou (1948: 71), ‘ou comme
ses (hommes) servent le chef.’
KG 1,5.4--5: vyatisikta bhavanti / vyatisikta iva va ime prana atmanam bhufüjantiti.
In his commentary, Udaya (1976: 25) first explains that iti (. . . bhufijantiti) i is used here in the sense
of ‘because’ (iti (abdo hetvarthah), then equates the breaths spoken of with sight etc.: yasmad ime
paridrsyamanas caksuradayah pranah susambaddhd eva santa atmanam sariram bhuñjanti palayanti
. palanam nama svocitatattatpravritihetutvam/sambandhas canyonyam sahayyacaranam/na hi
caksuradinam anyonyavartanabhijfiatve tannimitta sarirapravrttir upapadyate. As can be seen, Udaya
once mote glosses bhufijanti with palayanti. He also explains that the act of taking care of, serving
(palanam) consists in being the cause for the body to enter into activity as appropriate to each sense
and sense organ. It is not possible that the body engages in any activity of perceiving conditioned by
these unless they have separate spheres recognized as such; and their being related consists in their
helping each other. It is not necessary to take up an alternative interpretation which Udaya gives.

21 See recently K. Hoffmann (1967: 96 with note 196). I do not discuss here the optative bhujema,
used with ma, and the associated first singular form bhojam, both used in the Reveda, since at the
moment I do not have anything definite to contribute to the problems associated with these: they
are treated by Hoffmann (1967: 95—96).

7 edhyapita ye gurum nadriyante vipra vaca manasa karmana va/yathaiva te na guror bhojanivas
tathaiva tan na bhunakti srutam tat ‘As Brahmanas who, when they have received traditional instruc-
tion, do not respect their teacher in word, thought, or act are not worthy of having the teacher give
them food, so does the learning they have received not serve them.’ The verse is cited in Nirukta 2.4.

3 Durga (Anandasrama Sanskrit Series 88.1: 142 (Pune 1921)): na palayati na bhunakti/$rutaphalena
na samyunak tity arthah.

7^ The rest of the text involves difficulties not directly pertinent to my discussion.

25 As vardhayati is later connected only with vardhate, since the factitive vérdhati is eliminated. But
bhunakti semantically distinct from bhojdyati is not eliminated.

% It does not matter to my discussion that fypyati in the Rgveda is construed with a genitive, not an
instrumental (cf. Delbrück 1888: 133).

27 See PW V.301. Note that the use of either an instrumental or an accusative form signifying an object
consumed in constructions with bhuj persisted for a considerable time. Both constructions are used
by Patafijali in his Mahabhasya (Pune 1962—1972; e.g., Bhasya I1.163.2—3 (on Astädhyäyi 3.3.193):
abhoksyata bhavan mamsena yadi matsamipa asisyata ‘You would eat meat if you were to sit near
me’, 1.326.20 (on 1.4.24): ... odanam bhunkte ‘... is eating rice’, the first with the instrumental
mamsena , the second with the accusative odanam.

28 See section 1. Within the linguistic system in question, moreover, one cannot link bhunkté and
bhundkti by the semantic criterion which generally determines whether a verb that partakes in this
opposition has middle or active inflection, namely whether or not the results of the act signified by
the base is intended for the agent of that act (Panini 1.3.72: svaritafüitah kartrabhipraye kriyaphale,
1.3.78: sesat kartari parasmaipadam). The best known example is ydjate, said of a Yajamana, one
who serves as patron of ritual acts and for whom the fruits of these acts are intended, opposed to
ydjati, used of an officiant who performs such acts for the benefit of the Yajamana. bhunkté ‘enjoys,
uses, consumes' and bhundkti ‘aids, serves, protects’ are not related in this manner, and Panini treats
the verbs yaj, bhuj accordingly: In his dhatupatha, yaj is given with a svarita vowel as a marker
(yajA), so that atmanepada and parasmaipada affixes follow this under the semantic conditions
noted, but bhuj is listed with a high-pitched vowel marker (bhujA): Astadhyayi 1.3.66 then lets this
base take atmanepada affixes when it is used in particular meanings.

References

Bóhtlingh, Otto — Rudolf Roth, 1855-1875. Sanskrit-Wörterbuch (St. Petersburg: Kaiserl. Akad. der
Wissenschaften).

Debrunner, Albert (— Wackernagel, Jakob), 1954. Altindische Grammatik 2 (Göttingen: Vandenhoeck
& Ruprecht).

Delbrück, B., 1888. Altindische Syntax (= Syntaktische Forschungen 5) (Halle: Niemeyer).

Eggeling, Julius, 1882--1900. The Satapatha — bréhmana (Oxford: Clarendon Press).

Hoffmann, Kar, 1967. Der Injunktiv im Veda (Heidelberg: Winter).

Keith, A. Berriedale, 1914. The Veda of the Black Yajus School entitled Taittiriya Sanhita, translated
by A. Berriedale Keith (7 Harvard Oriental Series 18) (Cambridge, Mass.: Harvard University Press).

Mayrhofer, Manfred, 1963. Kurzgefaßtes etymologisches Wörterbuch des Altindischen 2 (Heidelberg:
Winter).

Monier-Williäms, Sir Monier, 1899. Sanskrit-English Dictionary (2nd edition) (Oxford: Clarendon Press).

Renou, Louis, 1948. Kausitaki Upanisad (Paris: Adrien-Maisonneuve).

Sahkarananda, 1968. The Kausitaki-Brahmana-Upanisad with the Dipika commentary of Sankarananda,
edited with an English translation by E.B. Lowell (Reprint) (Varanasi).

Schröder, Leopold von, 1900. Kathaka. Die Samhitd der Katha-gäkhä, 1. (Leipzig: Brockhaus).

Sreekrishna Sarma, E.R., ed., 1968. Kausitaki-brahmana (Wiesbaden: Harrassowitz).

Sreekrishna Sarma, E.R., ed., 1976. ‘Kausitaki-brahmana 2: Vyakhya of Udaya (Wiesbaden: Harrasso-
witz).

Vira, Raghu, 1968. Kapisthala-katha-samhita (Delhi).

Whitney, William Dwight, 1905. Altharva-Veda Samhita, translated by W. D. Whitney (= Harvard Orien-
tal Series 7 —8) (Cambridge, Mass.: Harvard University Press).

Who did discover the law of the palatals?

Neville E. Collinge (University of Manchester)

During the 1870s it suddenly occurred to people that two problems of Indo-European
history were each the solution to the other. The first was a Proto-Indo-European question:
had the ‘Ursprache’ a basic system of five vowels (as Greek or Latin — or Gothic, once its
spelling was understood — and even Celtic seemed to suggest)? Or was it of three, with only
/al as a non-high segment (as Indo-Iranian offered, and in a quite ‘natural’ way, in view of
the Semitic system)? The other nuisance was inside Sanskrit, which showed palatal stops as
reflexes of inherited dorsals where a velar articulation was expected. Scholars were inured
to the complexities of the dorsal series, where the skewed pairings of segment-types in the
Indo-European languages argued for either three types in Proto-Indo-European, or else two
there with collapsing of these (plus phonemicizing of variants of one of them) in one group
of languages: on which the battle still rages. But Sanskrit had a secondary palatalization
(with a different actual result (c not $) in the voiceless case). There is awkward distribution
too: forms like vacas and pacanti seem to palatalize /-ko-/, against the testimony of cakara
with /keko-/. The solution, once stated, was unassailable: Sanskrit also would argue for a
mid-front /e/ in PIE (and so for /o/ also), were it only granted a prehistoric /e/ of its own;
and if it were, the condition for its -ca- effect was honoured, and pacanti etc. could be seen
for the analogical products which they are. (Paradigmatic analogy is a splendid tool for ex-
plaining exceptions in a known output, but it is awfully stifling when the rule is not known
and the exceptions cannot be seen. Again, Sanskrit was the oldest witness, and so indeed
until Hrozny unveiled Hittite in 1917; but this solution showed it to contain derivative ex-
plicanda. Hence the delay and the sadness of some — Curtius, for example; cf. Jespersen
1922: 91 — at the demoting of this prime witness.) So at last emerged ‘the law of the
palatals’, which might as suitably have been titled ‘the law of the Indo-European vowels’.

Still, as we have learnt from Hoenigswald in so salutary a fashion, an unconditioned merger
(‘the central type of sound change’) is a handier diachronic affair than a complicated pro-
cess of splitting. So the thesis that /e, a, o/ became /a/ in Indo-Iranian should always have
been the horse to back. Perhaps palatalizing before /i/ in Indic would have commended it;
but apaciti- (Ved.) and cid (but kim, from MIA) practically exhaust the cited evidence. In
fact, the shifts are inside Indic and are: (e.g. *(—)k”e(—)>ke>ce>ca, with paradigm-levell-
ing at the last or next to last stage. But who said all that, or conceived it, first? Jespersen
(1922: 90 fn. 1) thought it futile to enquire; and two of the possible claimants themselves
deplored the attempt. But investigation is fair in an area where information has been leaked
piecemeal and interpreted loosely. At least one should gauge what credit is due, or not, to
various contenders and what testimonia might still be rummaged for.

Some of the Indo-Europeanists of the period can be readily eliminated from contention for
the ‘Prioritàtsrecht’. Two were involved simply as triggers. In 1870 Ascoli, having led the
way among those who (against Fick) reconstructed three basic dorsal stops in Proto-Indo-
European, opted for even a fourth series to solve the Sanskrit riddle. This seemed to touch
a nerve; it was seen to flout economy and chronology and never won hearts as the pro-
posed trio of original dorsals still does. Pedersen in 1924 (cf. 1962: 281) spoke roundly of
Ascoli’s ‘mistake’. Some explanation was urgently needed. Then in 1874 Amelung repeated
the suggestion (which was already in Benfey 1837) that Proto-Indo-European and Greek
/e/ had collapsed with /a/ in Sanskrit; but he showed more precision and a very relevant
concentration on /e/ alone. Amelung does not see the crucial link with the palatalized
dorsals, nor credits anyone else with it: so 1874 is the terminus a quo for the discovery. And
by 1879, with the treatments of Collitz and Hübschmann published in that year, the thing
was quite codified. We have our limiting quinquennium.

To those set aside (Benfey, Ascoli, Amelung) we can add others who contributed nothing,
or only on the array of basic velar consonants (Fick, Bezzenberger, Bugge). Some ten
candidates remain, and even of these five can, with a trifle more trouble, be winnowed out.
From the ‘document in the case’ #3 — of which more below — it emerges that news of the
law came quite fresh to Leipzig in 1876 and was so received by Osthoff and others, four of
whom are named by him. Three were Brugmann, Hübschmann and Leskien. None of these
is ever put forward as the discoverer; and only Hübschmann and Osthoff discussed it in
print at the time and both assigned credit elsewhere (in fact to Verner). The other named
recipient of the news was Saussure.

Now Saussure is said by Szemerényi (1964: 4 fn. 6) to have been 'very near to discovering
the law" (but cf. Szemerényi 1967: 68 fn. 1); and Streitberg (1913; cf. 1915: 205f.) had
written of him as among its discoverers (as also does Koerner as recently as 1978: 199).
Indeed, Saussure makes a claim for himself: *il me vint l'idée . . ." (1879: 369). Saussure
was undoubtedly precocious (cf. Pedersen 1962: 284 fn. 1; he was “der friihreife Schwei-
zer" for Brugmann 1909: 219); still he was only 18 at the critical time in Leipzig. Had he
revealed then that the solution was already in his mind (in October 1876) — and if he did
not then his claim fails — he must have provoked comment or challenge from those quick
wits. There was his remark to Streitberg in 1903 (see the latter's 1915 notice) that he had
not heard Verner's news in Leipzig; and on 21 July 1877 he read a paper (which was
published in 1879) which antedates the publishing of the law by Collitz and Osthoff. But
those scholars must have been working on the material months before. Saussure's famous
Mémoire (of December 1878) -- the great ‘prelaryngealist’ paper — combines perspicacious
morphology, as on the Proto-Indo-European nasal infix verbs, with really obscure, if ‘an-
regend’, phonology. As to the 1879 publication on the palatals business (given as a paper in
1877), Saussure admits (1879: 1 fn. 1) that his essay is overcondensed and short on data
and coherence; only two pages are ad rem. It is suggestive that he never repaired these
faults as promised. He wants a single non-high Proto-Indo-European vowel /a/ (1879: 362)
but with various “puissances” (1879: 363), with no phonological motivation. So we are
less far forward than was even Amelung in 1874. Four Brugmann-like variants are offered:
A, A5; a, a5. The first two respectively relate to some cases of i and à in Sanskrit and some
‘European’ cases of o (A), and to Brugmann’s law output (45). The second pair answer in
turn to ‘European’ /e/ and to most cases of /o/. True, cakara is explained as from
*k yak yA pra, so that the velars are affected by the following vowel; but when arca- vies

with arka- (etc.) we hear that arcati has /-ka-/ but arcami has /-ka9-/ (1879: 369). This
should mean that they share frontness, and that is untrue; analogy is not mentioned and
even the frontness feature is left inexplicit. Even Streitberg (1915: 206) concedes that here
is error and failure. In the second syllable Greek épos is assigned a? but kos A, which is
ad hoc. If these are insights, they are provokingly dark; if this is pioneering, it is rather late.
(And on the derivative nature of much of Saussure’s work, cf. Collinge 1979: 209; Koerner
1978: 149.) So the entire Leipzig group may stand aside.

We thus retain a short-list of five:
Collitz, Hermann (1855—1935)
Schmidt, Johannes (1843-1901)
Tegnér, Esaias Henrik Vilhelm (1843—1928)
Thomsen, Vilhelm Ludvig Peter (1842—1927) and
Verner, Karl Adolf (1846—1896);
— and we have five crucial ‘documents in the case’:
#1 = Schmidt 1881; #2 = Collitz 1886 (& 1887); #3 = Osthoff 1886; #4 = Verner 1886;
#5 = Thomsen 1920.

The law was first published in 1878. In Collitz 1878, which appeared in May, it is really
only presaged on the last page (305); and this is made good the next year. Osthoff 1878
sets out the law, and gives credit to Verner for having had the notion ‘1% to 2 years”
before; but Collitz is heavily ironic (1879: 207 fn. 1) over Osthoff's failure to find a copy
of his own earlier article; and (if Osthoff must prize thought over print) Collitz claims to
have ventilated it all verbally in the summer of 1876 with Fick, Bezzenberger and Benfey.
Of this there is corroboration (in BB 12 (1887): 248). But even so, it is a shame the date
Collitz quoted was not a year earlier, as we shall see.

‘Document #1° was submitted to KZ in June 1879, and Schmidt says he was writing it in
the spring of 1878. It links Indo-Iranian palatalization to following front vowels. But in
KZ for 1877 (vol. 23) Schmidt had an article siding trenchantly with those who viewed a
Proto-Indo-European /e/ — or even a common European vocalism — with scepticism. Such
a volte face within a year is remarkable, if unprompted. Even more remarkable is his
writing (1881: 63) that he could well have claimed the inventor's right and is constrained
from doing so only by not having the time or inclination to deploy the necessary Avestan
and Vedic evidence (and by one or two troublesome exceptions); but he has been in-
cluding the law 'since May 1877' in his lectures — a phrase to bear in mind. He does not
want an unfruitful combat over the title of discoverer, and anyhow Thomsen of Copen-
hagen ‘probably had the idea earlier than all of us'. That sounds as if counterclaims alone
were restraining him. He insists on the second-hand status of Hübschmann's and Osthoff's
knowledge in the affair, and declares that he had written his version before the publications
of Collitz or Hübschmann or Saussure. But the phrase ‘May 1877’ is a curiosity. He had,
according to ‘document #3’, been told of the law in January 1877; so the remark is off the
point and shows rather that he took a longish time to assimilate the implications. (Four
months is a longish time in a matter where Collitz censures Osthoff for being two months
late.) Finally, nobody but Schmidt supports Schmidt's claim (implied even in 1884).

‘Document #2° was written by Collitz in 1885. As he was in many ways a Fick man and
Schmidt was a pupil of Schleicher it is intriguing to find here some German solidarity —
but it is for outflanking the Leipzig school that Schmidt wins Collitz's praise. The article

concedes, however, that Thomsen, Verner and Saussure may be independent discoverers
of the law. We have considered Saussure; the others remain possible. Tegnér is also men-
tioned as a co-inventor (the news comes from a letter from Thomsen, which speaks of the
law as ‘lying in the air' — a recurrent phrase in this affair). So Collitz seems to forgo his
claim (and, by publication priority, a fortiori to disclaim the honour for Schmidt); his con-
versations of 1876 fade into nothingness. Yet Prokosch, always an encomiast of Collitz,
gives him credit for this law (1939: 93) and is followed by Porzig (1954: 27); more arrest-
ing is that Allen (1973: 123) should think Collitz ‘an independent discoverer of the law’.
This judgement may have some colour.

Then came ‘document #3’, by Osthoff, who writes in fury over the stealing of Leipzig
thunder by Collitz and Schmidt. The nub is that Osthoff and others learnt the news of the
law from Verner “at the beginning of the winter semester of 1876” (in fact, October 1876)
— and the major names are given of those who received this bolt from the blue. Later,
writes Osthoff, in January 1877 he visited Schmidt in Berlin and passed on the news, which
Schmidt seemed to recognize as revolutionary and destructive of many existing explana-
tions of his own as to Indo-European matters; he became pensive (“nachdenklich”). But
later he referred to the disclosure (181) as if it had not come via Osthoff; and even in 1884
ostentatiously resigned his claim so as virtually to leave a toe in the door. For Osthoff the
law is Verner’s, and is twice so characterized; resisting the temptation to organize a collo-
quium on the subject, Osthoff set it out in print in 1878, with Verner duly honoured.

Verner himself produced ‘document #4 in November 1886. It appeared in a journal, and
in a year, wherein Collitz and Schmidt are also named as the inventors. Rooth’s reprint of
the text (1974: 17--20) is very helpful, as was Szemerényi’s reminder of Osthoff's state-
ment (1967: 68 fn. 1). Verner starts to the effect that for eleven years he has given up such
claims to the inventorship as he once had. Well, then he had them. At the end he lyrically
exclaims that the solution had been an over-ripe fruit on the tree of philology and after its
fall the only credit is to have shaken the tree hardest. So he agrees with Thomsen that the
personal question is neither here nor there, and an orphanlike title for the dubious child is
best; enquiry into paternity, as the Code Napoléon said, is forbidden.

Yet things said in the centre of the document are rather curious. Verner says he discussed
the matter with Thomsen in Copenhagen in the summer of 1875 (when, one must remem-
ber, Collitz was only 19). Now that spring and early summer Verner was at Aarhus recu-
perating, and working on those two famous papers which appear back-to-back in KZ 23
(dated 1877, but the early papers are said to have appeared in 1876). Of these the first sets
out Verner's own famous law, which involves him in handling unfixed word accent in Indo-
European. The second treats of the implications of his law for the theory of ablaut. That,
of course, is the natural outcome of interest in Proto-Indo-European accent positioning,
and should lead him on to feeling a need for a mid-front and a mid-back vowel in Indo-
European (if ablaut was then taking the shape in his mind which it later took in theory).
When, however, he found that Thomsen had independently come to the same result by
another path, he had no doubt that Thomsen deserved recognition as the inventor (and
hoped and expected that Thomsen would write a paper on the subject). This is less straight-
forward than it appears. Thomsen's paper is printed in his collected works (the second
volume of which came out in 1920). It certainly states the law; it does so, however, on the
basis of control by contextual variants of a single low vowel; and although it is consider-
ably tidier it is not substantially different from the approach of Saussure (or of Brugmann

on Indo-European vocalism). There is absolutely nothing of the dynamic inter-theoretical
linkage of which we might well believe that Verner was capable. True, one is merely
surmising that, having gone from his own law to a discussion of ablaut, Verner had mental-
ly pursued ablaut theory to the point of revising the Proto-Indo-European vowel inventory.
It is possible that he had not, and that it only occurred to him at the moment when he
heard what Thomsen, from another point of view, had to say. But what if Thomsen simply
established to Verner’s satisfaction that he (Thomsen) had been thinking about all this at
an earlier time? Then, for Thomsen to substantiate that, he would have to claim to have
been so thinking early in 1875. Yet we have seen that the terminus a quo of this whole
business is Amelung’s paper of 1874. So, all that Thomsen could effectively say is that he
thought of the law in ‘late 1874’ or ‘early 1875’. And that is exactly what in his own state-
ment he does say.

To continue, however, with Verner. He then reveals that he wrote a letter in August 1875
to the editors of KZ in which he reported both the essentials of the law and that Thomsen
already had some right to it (had, in fact, spoken of it to some audience or other) and was
about to write an article on it. Then, he writes, when he felt fitter and went to work in
Halle he began to visit Leipzig; and one evening in October 1876 was involved in an animat-
ed philological discussion in a café. His contribution to the lively proceedings was a brief
outline of the law — including a mention of Thomsen’s name. He was upset when he came
to think ‘of what he had done; he realised that the thought once spoken was no longer a
secret: ‘one cannot take out patents in philology’, he sadly observes. He does not know
whether to blame the company for shrewdly seizing on to the idea at once, and beginning
its exploitation, or to blame himself for having destroyed Thomsen’s right of priority, as
Thomsen had not yet published. He consoles himself with the thought that at least he had
written the letter to XZ and so protected Thomsen. He allows that Tegnér, too, in Lund
had at least begun to write on the subject but was inhibited by a flood of solutions from
various sides. This seems to refer to some time before October 1876; but there are prob-
lems about taking that as Tegnér’s moment, to which we shall come.

Now comes ‘document #5’. Between 1919 and 1922 Thomsen published his collected
works, and in the second volume (dated to 1920) he finally printed his article on the law.
‘Document #5’ is its preface. In 1877 he had offered it to KZ, the first half being complete
and ready for publication and the second in little more than draft form; and this is how it
is given in 1920. It did not appear in KZ. On its submission Thomsen had a letter from
Ernst Kuhn, one of the editors, warning him off, because the solution was ‘lying in the air’
(or, we might say, ‘floating over Europe’) — a phrase which clearly pierced Thomsen.
Schmidt and Hiibschmann were said to be working on papers to the same effect. The
latter’s was published in 1879 and, as promised, mentioned Verner; Schmidt’s article is
presumably our ‘document #1’ (despite the author’s dating of its writing to 1878) — which
is ironic, as Thomsen’s could easily have been given to the world before it. But there are
mysteries here. How does it come about that when Ernst Kuhn receives an article on the
law from Thomsen in 1877 — admittedly about a year behindhand, because Thomsen
visited Oxford by invitation in 1876 and spent a year preparing his lectures in another field
and subsequently gathering them into a book — Kuhn responds as he does? Has he not
been told already in 1875, by Verner, to expect such an article to be offered by Thomsen,
as the law’s discoverer? Was the letter really mislaid, or just ignored? Why did not some-
body at KZ write to Schmidt and Hübschmann and tell them to hold back until it was

established whether Thomsen had quit the field or not? E. Kuhn offers an ‘explanation’
(1886: col. 1840) which shifts the blame without answering our questions: that the letter
should fail to pass between the Kuhns over several years (and yet not enough years) is
eyebrow-raising. (The editors of KZ in 1877 were, by the way, the two Kuhns, Leskien and
Schmidt.) On the face of it, Thomsen was not well served; his claim to the discovery is not
shaken.

No, the disappointing thing is his path to the discovery, which is so much less interesting
than the exploitations by the Germans (who linked it to ablaut theory), and than what
Verner might have made of it. And there are other oddities about Thomsen’s statement. He
speaks of agreeing with Verner (referring to “document #4’) in leaving unedifying disputes
about right of priority to the German scholars. He also reports meeting Verner in the
summer of 1875; but in the many detailed letters which Verner wrote to him about that
time, he says, the law was not mentioned at all; and in the conversation in the summer of
1875, when they did discuss it, Verner never once suggested that he himself had made the
discovery. This would seem to make Thomsen’s case very strong indeed. Yet there is the
rather elastic way in which he declares that he had thought of it all ‘towards the end of
1874 or at the beginning of 1875’; although one might well not manage to place an idea to
a month or two, and particularly if it comes out of one’s lectures in a given academic
session. More worrying is the fact that Thomsen goes on to state that his results were im-
mediately afterwards put forward in lectures: “mine resultater har jeg straks derefter
fremstillet i forelaesninger" — a sentence which echoes Schmidt’s disingenuous remark
of 1881.

Thomsen also gives (in 1920; not in 1902) credit to his friend Tegnér who ‘at the same
time as myself had come to the same conclusion’. Tegnér’s part is hard to assess. He spent
most of his career as a biblical scholar or as an adviser to the Swedish Academy on diction-
ary matters, and he only made one written foray into Indo-European philology. This ap-
parently exists in a forty-page version, bearing directly on the law of the palatals, but is an
elusive monograph. The full form, dated to 1878, is, however, unimportant; not only
because of the lateness of the date but because we know (e.g. from ‘document #4’) that
what was available in Lund at the relevant time, by Tegnér, was a printed fragment (sic) of
five pages with a gummed-in sheet at the end explaining why the work was broken off.
Apparently, what Tegnér wrote he wrote after the disclosures of October 1876 in Leipzig
and January 1877 in Berlin. He lectured on the discovery in 1877 (see Mayrhofer 1983:
139). What and when his own earliest thinking was, and whether he was in mental com-
munion with Thomsen in 1874—5, we cannot know. At best, we can allow that if Thomsen
is the discoverer (despite a disclaimer at 1927 (1902): 94, which gives 1875 as the earliest
relevant date), some prompting came from Tegnér. Yet must we leave the discovery in
Thomsen’s hands? Let us return to Verner.

Verner was modest and non-combative; but he was more interesting than that. He had
something of the characteristic that the Scottish call being ‘fey’, of being subject to sudden
almost miraculous flashes of insight, of putting two and two together where others do not
see that two and two are even there. Remember the famous story of how he discovered the
essential feature of his own law. He reminds one of Thomas Edison, who was no great
scientist but who frequently realised what might be put with what to produce an invention.
It is credible that the putting together of a problem of German voicing, Indo-European
mobile accent, Indo-European ablaut phenomena, Indo-European cardinal vowel location —

that the connecting of these in a way that the pedestrian approach of Thomsen does not
would be very much in Verner’s line.

There is more. The law of the palatals has three interesting characteristics. First, of all the
‘laws’ of Indo-European (on which see Collinge 1985) only two were allowed at once to be
correct and have never been overthrown or even seriously doubted. The law of the palatals
is one, and Verner’s law is the other. Secondly, secondary palatalization in Sanskrit looks
like a body of inexplicable exceptions to normal reflexes. Verner made his name by his
famous explanation of apparently inexplicable exceptions to the great regularity which we
have long called Grimm's law. Thirdly, the law of the palatals takes a visible problem in
Sanskrit and solves it by adducing critical evidence, the incidence of a front vowel, which is
as such visible only in other languages. This is exactly Verner's technique in his own law,
where he explains Germanic consonant forms (in terms of their voicing or non-voicing) by
the positioning of Vedic, and hence Proto-Into-European and pre-Germanic, word accent.
Frankly, if we could find somebody who knew everything about 19th century Indo-Euro-
pean studies and scholars but who had somehow not heard of the law of the palatals as
such, and if that person were told of the law and asked to name a scholar most likely to
have invented it, I am pretty certain he could only name Verner.

To summarize: Collitz, Thomsen and Verner remain in line for honour. Collitz did not
discover the law first. But his clear, correct and courageous view of the Proto-Indo-Euro-
pean vowel system, in which one must have the guts (he says) to posit proper mid segments
and which not even the laryngeal theory really disturbs, combines with his unerring linkage
of the two problems in 1879. He had got there by early 1878, and nobody proves that he
must have known the essentials of Scandinavian thinking before that. That is, he would
have reached the same right result had no other done the same; and Allen’s judgement
needs only to be amended to ‘a potential independent discoverer of the law’. Of the others,
maybe Thomsen did get there first (in 1874 — or early 1875), but without a credibly oper-
ational vowel phonology and with some prompting from Tegnér; and then Verner's notions
slipped into place as a result of Thomsen's conversation. Or else Verner had arrived at the
solution in mid — 1875 but, faced with Thomsen's chronological claim, had no stomach for
a dispute or even for writing it all up. That would explain both his inability to check his
personal enthusiasm in the Leipzig café and his embarrassment thereafter. For Osthoff
(1886: 20) it is, finally, the law of Tegnér, Thomsen and Verner (with alphabetical tact).
But at least until the chips are down — and with so many talks unrecorded and letters
untraced the chips must stay up, probably for ever — this writer would give it to Verner.

References

Allen, W. Sidney, 1973. “X8wv, ‘ruki’, and related matters". TPS 1973: 98—126.

Amelung, A., 1874. "Erwiderung", KZ 22: 361—71. `

Ascoli, G.I., 1870. Lezioni di fonologia comparata {= Corsi di glottologia 1) (Turin and Florence:
Loescher).

Benfey, T., 1837. “Review of Pott’s Etymologische Forschungen auf dem Gebiet der indogermanischen
Sprachen”, in: Ergänzungsblätter, Hallesche Allgemeine Literaturzeitung, 114--117, cols. 905--33
(= Kleinere Schriften, edited by A. Bezzenberger 1890 (Berlin: Reuter) 1.2.)

Brugmann, Karl, 1876. “Zur Geschichte der stammabstufenden Declination”, in G. Curtius, Studien zur
griechischen und lateinischen Grammatik, (Leipzig: Hirzel), 361--406.

Brugmann, Karl, 1909. “Obituary of Osthoff", ZF 24: Anz. 218—221.

Collinge, N.E., 1979. “Review of Current trends in linguistics 13 (edited by Thomas A. Sebeok)”,
Language 55: 207--11.

Collinge, N.E., 1985. The Laws of Indo-European. (Amsterdam & Philadelphia: Benjamins.)

Collitz, H., 1878. "Ueber die Annahme mehrerer grundsprachlicher a-Laute”, BB 2: 291-305.

Collitz, H., 1879. ‘Die Entstehung der indo-iranischen Palatalreihe”, BB 3: 177—234.

Collitz, H., 1886. “Die neueste Sprachforschung und die Erklärung des indogermanischen Ablautes”,
BB 11: 203—42. (Also, 1887, ibid. 12: 243—48.)

Curtius, G., 1876. Studien zur griechischen und lateinischen Grammatik (Leipzig: Hirzel).

Hübschmann, H., 1879. “Iranische Studien", KZ 24: 322—415.

Jespersen, O., 1922. Language: its nature, development and origin (London: Allen and Unwin).

Koerner, E.F.K., 1978. Toward a historiography of linguistics: selected essays (= Studies in the history
of linguistics 19) (Amsterdam: Benjamins).

Mayrhofer, Manfred, 1983. Sanskrit und die Sprachen Alteuropas (Nachrichten der Akademie der Wis-
senschaften, Göttingen). (Göttingen: Vandenhoeck & Ruprecht).

Osthoff, H., 1878. “Formassociation bei zahlwörtern”, in: Morphologische Untersuchungen 1, edited
by Osthoff, H. & K. Brugmann. (Leipzig: Hirzel), 92--132.

Osthoff, H., 1886. Die neueste Sprachforschung und die Erklärung des indogermanischen Ablautes (Hei-
delberg: Petters).

Pedersen, Holger, 1924. Sprogvidenskaben i det nittende Aarhundrede: Metoder og Resultater (Copen-
hagen: Gyldendalske Boghandel), translated 1959, ?1962, by J.W. Spargo (Bloomington: Indiana
University Press).

Porzig, W., 1954. Die Gliederung des indogermanischen Sprachgebiets (Heidelberg: Winter).

Prokosch, Ernst, 1939. A comparative Germanic grammar (Baltimore and Philadelphia: LSA).

Rooth, E., 1974. Das Vernersche Gesetz in Forschung und Lehre (Lund: Gleerup).

Saussure, Ferdinand de, 1879. “Essai d’une distinction des différents a indo-européens", MSL 3: 359—
70 (= Recueil des publications scientifiques de Ferdinand de Saussure, 1922: 379—90).

Schmidt, J., 1877. ‘Was beweist das e der europäischen Sprachen für die Annahme einer einheitlichen
europäischen Grundsprache?”, KZ 23: 333—75.

Schmidt, J., 1881. “Zwei arische a-Laute und die Palatalen”, XZ 25: 1-179.

Schmidt, J., 1884. “Address”, in: Sitzungsberichte der Königlich Preussischen Akademie der Wissen-
schaften (Berlin). See esp. 741.

Streitberg, W., 1913. “Obituary of Saussure”, IF 31: Anz. 16.

Streitberg, W., 1915. F. de Saussure ZJb 6: 2.203-13; reprinted in Portraits of linguists, edited by T.A.
Sebeok (Bloomington: indiana University Press), 1966: 100—110.

Szemerényi, O.J.L., 1964. *Structuralism and substratum", Lingua 13: 1.29.

Szemerényi, O.J.L., 1967. “The New look of Indo-European: reconstruction and typology”, Phonetica
17: 65--99.

Tegnér, E.H.V., 1878. De ariska sprakens palataler (= Arsskrift, 14). (Lund: University Press).

Thomsen, V.L.P., 1920. “Der arische a-Laut und die Palatale”, in: Samlede Afhandlingen 2: 303—27
(with preface) (Original 1877).

Thomsen, V.L.P., 1927. Geschichte der Sprachwissenschaft bis zum Ausgang des 19. Jahrhunderts,
translated by H. Pollak, with permission and revision by Thomsen (Halle: Niemeyer). (Original 1902,
reprinted 1919 in: Samlede Afhandlingen 1).

Verner, K.A., 1886. “Zur Frage der Entdeckung des Palatalgesetzes’”, in: Literarisches Centralblatt cols.
1707--1710. (= Rooth, E.: Das Vernersche Gesetz in Forschung und Lehre (l.und: Gleerup), 17—
20). (See also 1903 Afhandlingen og Breve (Copenhagen) no. 53.)

The second plural of the Umbrian verb

Warren Cowgill + (Yale University)

1. Many of the verbs in the Iguvine Tables are singular active imperatives with the ending
-tu, -tu, e.g. enetu lal = enetu Vlal.* The etymology and value of this ending are clear. It
corresponds to Oscan -tud in likitud etc., Early Latin -t0d in datod etc., Classical Latin -t0
in dato etc., Greek third singular -rw in Gorw etc., Vedic -tat in bhavatat etc., and can be
confidently reconstructed as Proto-Indo-European *-tod. (Whether this is originally an ab-
lative of the pronoun */ó-, as stated e.g. by Brugmann (1916: 572), is another matter,
about which I have the gravest doubts.) Following recent usage, I will call imperatives made
with the element *-tod or in analogy to it ‘imperative 2’, in contrast to ‘imperative 1’,
which is formed without *-töd. /

In Umbrian, as in Latin, the singular of imperative 2 functions as both second and third
person. Second person is clear in the prayer texts, e.g.

V]Ia30 futu fos pacer pase tua
‘be favorable [and] propitious with thy peace’.

Third person is clear in some of the instructions for performance of the rites described in
the Tables. Nussbaum (1973) has adduced good reasons for thinking that the rituals de-
scribed in Tables I and VI—Vila were originally in the second person, and that this is
preserved in Tablet I, while the redactor of Tablets VI and VIIa intended to transpose into
the third person, but did not always succeed. A clear example where the change in person
is signaled by a change in pronouns, while the verb itself remains unaltered, is Ib13 steplatu
... tefe ‘require ... for yourself? vs. VIb51 stiplatu ... seso ‘he shall require . . . for him-
self”.

2. The corresponding mediopassive imperative 2 ending is mu, -mu, e.g. the very frequent
pe(r jsnimu, persnihmu, pe(r)snimu, persnih(i)mu *precator'. Here also the person is ambig-
uous between second and third, and in Tablet I persnimu etc. are second person and in Tablets
VI and VII third person. This ending isclearly the Umbrian analog of the Old Latin deponent
imperative 2 ending -mino, e.g. progredimino Plautus, Pseudolus 859. Latin -mino (or
already *-minod) was clearly formed to the second plural mediopassive indicative and impera-
tive 1 -mini on the pattern of the active second and third singular imperative 2 -to(d) beside
second plural imperative 1 -te (and indicative -tis, *-tes). The rule would be ‘form the singular
of imperative 2 by replacing the nucleus (and coda, if any) of the final syllable of the plural

* It is a great pleasure to thank Michel Lejeune for reading and commenting on a preliminary draft of
this paper.

of imperative 1 by -o(d/'. In Oscan censamur “he shall be assessed’, Tab. Bant. 19, the voice
has been redundantly marked by replacing the -d of *mud with the -r of other mediopas-
sive forms such as uincter ‘is convicted’; cf. the -minor of late Latin grammarians, Leumann
(1977: 573).

Umbrian -mu and Oscan -mur are indirect evidence that corresponding to Latin -mini,
Osco-Umbrian had second plural mediopassives with an ending of the shape *-mV/C); direct
evidence are the mediopassive plural imperatives 1 -mo, 2 -mumo to be discussed in sec-
tions 3. and 4. Of the various explanations offered for Lat. -mini, the best in my opinion is
the old one, that it is originally the nominative plural masculine of a participle in *m Vno-
(or *-mno- with anaptyxis?), so that Lat. sequimini is originally truncated from sequimini
estis/este (: &róuevot &aore, ënduevor éore), used in place of the inherited second plural me-
diopassive ending Proto-Indo-European *-dhwe or *-dhwo which is found in Skt. -dhve,
-dhva(m), Hitt. -ttuma(-), Luw. -tuwa-, and which probably underlies Gk. -ode. If so, Oscan
‘mur and Umbrian -mu, -mo, -mumo imply for the immediate ancestor of Osco-Umbrian
mediopassive participles in *-mo-, of the type seen in East Baltic, Slavic, and Luwian (e.g.
Lith. mylimas ‘loved’, Latv. jutams ‘felt’, OCS iskuSaemu ‘being tempted’, Luw. a-i-ia-am-
mi- ‘made’). Whether these represent a formation distinct from the -m/V)no- of Lat. -mini,
femina, alumnus, Gk. -uevo-, Av. -mna-, Skt. -mana-, Toch. B -mane, A-mam, OPru. po-
klausimanas ‘heard’, or are all due to independent sound changes of *-mn- to -m- is a prob-
lem I will not go into here."

3. The Umbrian plurals of imperative 2 are active -tutu (Ib: 6 times), -tuta (HI: 5times),
-tuto (VIb, VIla: 12times), and mediopassive mumo (VIb, Vila: 5 times). The active forms
here no doubt are the Umbrian analog of Lat. -fofe, i.e. are made by adding the active
plural ending of imperative 1 to the active singular of imperative 2, on a pattern analogous
to that of Latin ama : ama-te = amato : X; X = amato-te. (This was seen already by Brug-
mann (1892: 1327). Umbr. -mumo is the mediopassive pendant to this, i.e. is made by
adding the ending of mediopassive plural imperative 1 to the mediopassive singular ending
mu of imperative 2 (as if Latin progrediminò had a plural $progrediminomin1).

As has often been noted, although the pluralizing elements of Umbr. -tuto, -mumo are
etymologically second person only, these plurals are used also as third person, following the
double value of the singulars -fu, -mu, in contrast with Latin, where the active of impera-
tive 2 has a distinct third plural in -nto(d).

4. That the plurals of imperative 1 also are found in Umbrian was seen by Wackernagel
1907 (following a brief remark of Bücheler (1883: 197)). Etatu Ib 21, 22 = etato VIb 63
and armanu kateramu Ib 19f. » arsmahamo caterahamo VYb 56 are in a context that de-
mands imperative 1: these are commands addressed to the men of Iguvium to be carried out
at once, not instructions about how a rite is to be performed whenever the occasion for it
arises, or prayers to deities asking them to behave in a certain way to the people of Iguvium
or to their hostile neighbors. Wackernagel's insight was accepted by Brugmann (1911/1912
and 1916: 520, 577) and by Buck (1928: 362), and then forgotten until Rix (1976b: 225,
n. 20, 1976c: 162, n. 10) restored it to the attention it deserves.?

On the origin of the ending -to Wackernagel and Brugmann disagreed. The latter thought it
was from *-td, either *-t 7, i.e. the plural imperative ending *-te elided before a particle *a,
or the dual ending *-td seen in OCS -ta and Lithuanian -ta, refl. -to-s (so first Brugmann

(1892: 1327)). Wackernagel (1907: 318f.) thought it was *-to, an o-grade of the second
plural ending *-te that appears e.g. in Gk. -re, OCS -te. Brugmann (1911/12) defended his
own view, Wackernagel (1911/12: 258f., n.2) replied, and Brugmann (1916: 577f.)
responded once more.

5. In my opinion, Brugmann's derivation from a dual *-ta@ is more likely both phonological-
ly and morphologically than Wackernagel’s derivation from a plural *fo. (Brugmann’s
alternative suggestion of a plural *-te plus particle *@ has nothing to recommend it.) Phono-
logically there is a difficulty for Wackernagel in the spelling -tuta of Table III in ustentuta
5, upetuta 10, etuta 11, fertuta aituta 13. While there is no serious difficulty in an ad hoc
assumption that Proto-Indo-European final *-o, of which there is no certain Umbrian ex-
ample outside these endings, had simply remained, as *o normally did elsewhere in Umbrian
(cf. e.g. von Planta (1892: 109) for examples in initial syllables, and von Planta (1892:
567) for a possible example of final *-o in supu IV 17), Wackernagel’s derivation requires
a further ad hoc assumption that final *-o had undergone a lowering in articulation such
that it could be written with (a) as well as with (u) in the native alphabet. But, as far as I
see, all other clear or possible cases of *o are written in the native alphabet only with (w,
. never with (a). Thus e.g. pusnaes la 2, pustnaiaf Ib 11 (cf. postne VIb 11); ustentu Ta 3 etc.
(: ostendu Via 20); ukriper la 5 etc., ukar Ib 7 (: ocriper Via 23 etc., ocar VIb 46);
puplum Ib 10 (: poplom VIla 15). To suppose that final #0 had undergone lowering to
[o] while non-final *o did not is to introduce a complication into Umbrian historical
phonology for which there is no evidence outside the second plural verb, and which has no
support in the general phonetic tendencies of Umbrian and Osco-Umbrian.

On the other hand, Brugmann's derivation from *-/d simply adds one more example for the
development of *4 to [o]? that is already well known for Osco-Umbrian in such categories
as the nominative singular of a-stems, the nominative-accusative plural of neuter o-stems,
and the active singular imperative 1 of-d-verbs. This sound is regularly written o in the Latin
alphabet, but both a and u in the native Umbrian alphabet, e.g. accusative plural neutrum
arvia la 3 etc., aruvia III 31, but arviu Ia 12 etc., aruio Vla 56 etc.; nominative singular
panta muta Vb 2 but entantu mutu Vb 6: imperative 1 stiplo Vla 2, aserio VIa 4. The
imperative plurals -tu, -mu, -tutu on Table I, -tuta on Table III, and -f0, -mo, -tuto, -mumo
on Tables VI and VII fit this pattern with no difficulty.

6. Morphologically also Brugmann’s derivation is superior. Nowhere in Indo-European
outside Umbrian is a second plural *-to clearly attested. In languages that keep *-e and *-o
apart, and where the second plural endings are not suspect of having dual origin, we find
only *-te, viz. Gk. -re; -te or its reflexes in all Slavic languages, even those with -mo in first
plural; Lith. -te, refl. -te-s:* and probably Paelignian imperative cite ‘ite’, perfect lexe
‘legistis’, both Vetter (1953 no. 213); although I can cite no example of *o in absolute
final in Paelignian, note e.g. iouiois puclois Vetter (1953: 202), suois cnatois Vetter (1953:
203), solois Vetter (1953: 214), where the diphthong oi in final syllable is preserved, in
contrast to the Latin development to -eis and then -is. Cf. also the Irish imperative type
taibrith Wb 6a 20, léicid Wb 6b 29, which on the evidence of the lenition after bad ‘be ye'
at Wb 24b 1 appears to come from plural *-£e rather than from dual *-tes. Note also Hitt.
-tten, -tteni, which, whether old plurals remodeled after first plural -wen, -weni or old duals
with -e- from plural *-fe in place of the *o implied by Gk. -rov and Toch. B third dual -tem,

normally have e, while the variant -ttani (a list e.g. in Kronasser (1965: 379ff.)) is more
likely the result of Hittite accent placement on a preceding syllable than a relic of Proto-
Indo-European (Proto-Indo-Hittite) ablaut. Although Tocharian A -c, beside which B -cer
has an additional -er of unclear origin, could be phonetically from *-tes as well as *-fe, I
see no reason to suspect dual origin for this ending, so that Tocharian counts as another
probable witness to second plural *-te. The Old High German second plurals beret, berat,
birit are a problem which I may treat elsewhere, but certainly do not require a Proto-Indo-
European second plural *-to (cf. Brugmann (1892: 1359; 1916: 625f.) for a number of old
suggestions about these). Armenian -yK* in e.g. pres. ind. sirek', layk', impf. sireik*, pres.
impv. mi sirek‘, pres. subj. siric 'ék*, aor. ind. sirec ek, sirec'ayk', aor. impv. sirec'ek*, aor.
subj. siresjik‘ certainly comes from *te(-) or *te(-), not *-to(-), although I cannot say
whether this is an old plural *-té remodelled with addition of -k', or an old dual *-tes, in

which case the -k* may perhaps be a generalized sandhi variant (original before *w-?) of
Ze

Nor is there any raison d’être for a variant *-to in this ending. In general, it is not licit to
posit ablaut variants of the same morpheme without either data too clear to be ignored, or
else a rationale by which the proposed ablaut clearly fits into a well established pattern
with well established conditioning. Neither of these cases exists here. Wackernagel’s appeal
to first plural -m0s, -mo beside -mes, -me (1912/13: 259) does not help, since the attested
first plural forms can be explained on the basis of primary *-mos, secondary *-me(-NJ, with
the vowels distributed like in nom. *ékwos, voc. *ékwe, and with subsequent local creation
of new variants -mes (Greek Doric -ues, poetic -ueoda) and -mo (Serbo-Croatian, Slovene,
Ukrainian, White Russian).

7. On the other hand, there are parallels for dual forms being generalized over plurals in
languages that are losing the distinction of the two numbers. Within the first and second
persons of the Indo-European verb, the clearest cases are probably the Latvian dialects with
-ma-s, -ta-s instead of -mie-s, -tie-s as reflexive plural forms, and the Polish and Czech dia-
lects with first plural -wa or -ma in place of standard Polish -my, Czech -me and second
plural -ża in place of standard Polish -cie; for all these see Vaillant (1966: 27f.). But also
the Old Irish second plural absolute of the type befi/the, ceste, gigeste seems to require a
Celtic ending *-tes, not *-te, and this *-tes cannot be explained as analogic to the second
singular in the way that Latin indicative/subjunctive -tis beside imperative -że can. It there-
fore seems to come from the Proto-Indo-European dual *-tes of Sanskrit second person
-thas, third person -tas, Avestan third person -to, -@0, Gothic second person -ts, Old Church
Slavic third person -że (perhaps with help from first plural *mos); cf. Cowgill (1975: 59f.).
If so, Latin -tis itself may well be from dual *-tes; and cf. section 6 above for the possibility
of dual origin for second plural endings in Hittite and Armenian.

Another possible example for generalization of dual over plural is the Hittite first person
plural endings, which are normally -wen, -weni (-wani, -uni) in active and -waSta(-) in
mediopassive, and begin with -m- only when the preceding sound is u. If Hittite comes
from a language which differentiated first plural endings in -m- from first dual endings in
-w-, as Proto-Indo-European Proper certainly did (for this term see Cowgill (1979: 25)),
then Hittite has clearly innovated by generalizing dual forms except after u. But, given the
likelihood that the Indo-Hittite hypothesis is correct (Cowgill 1979), it is possible that
Hittite comes from a protolanguage which had not differentiated first person plural from

first person dual by assigning different meanings to what were originally phonetically con-
ditioned variants.

Within languages which had already lost a distinction between dual and plural in first and
second person verbs, but retained distinct dual forms of first and second person pronouns,
Bavarian and Modern Icelandic have generalized dual forms over plural in these pronouns:
Bavarian only in second person es, enk, enka, icelandic in both, við, þið, etc., beside which
the old plurals vér, þér survive as honorifics; cf. Stefán Einarsson (1945: 68, 122). For
similar phenomena in Faroese and in Norwegian and Swedish dialects cf. e.g. Bugge (1855:
247, 254ff.).

8. Latin and Irish seem to have used the contrast of plural *-te and dual *-fes to fashion a
contrast between imperative and indicative, Lat. agite # agitis, Prim. Ir. *berete # *-beretes,
vs. Gk. impv. Aéyere - indic. Aéyere, secondarily using the old dual *-tes in all non-impera-
tive tense/mood categories, even though outside present indicative there was no direct
contrast with imperative — e.g. Lat. impf. agebátis, subj. agátis, perf. egistis; Prim. Ir. subj.
*.berates, fut. *gwigwessetes.

From the use of dual *-/2 in imperative, it appears that Umbrian had no such contrast, at
least in the active voice. And I suggest that the ending *-/2 is in fact found in Umbrian
outside the imperative in the future perfects couortuso VIb 64 and benuso Vib 64, 65.

Nussbaum (1973: 366ff.) has very persuasively argued that these are second person active
forms, not third plurals or passives (impersonals) as previously supposed, and proposes to
take them as singulars, of deponent shape. In taking benuso and couortuso as second
person, Nussbaum is undoubtedly correct, and has advanced greatly our understanding of
these forms. Herewith is the relevant passage, as I punctuate it: (VIb 62—65): Ape este
dersicurent, eno deitu “etato Iiouinur” porse perca arsmatia habiest. Ape este dersicust,
duti ambretuto euront. Ape termnome couortuso, sururont pesnimumo, sururont deitu,
etaians deitu. Enom tertim ambretuto. Ape termnome benuso, sururont pesnimumo,
sururont deitu etaias. Eno prinuatur simo etuto erafont uia pora benuso.

In literal, unidiomatic English I render this, When they will have said this [viz. the preceding
prayer], then he that will have the perca arsmatia shall say, ““Go(?), Iguvians”. When he will
have said this, the same ones (or, you same ones) shall circumambulate a second time.
When you will have returned to the terminus, they (or, ye) shall pray in the same manner;
he shall speak in the same manner, he shall say [that] they are to go(?). Then they (or, ye)
shall circumambulate a third time. When you will have come to the terminus, they (or, ye)
shall pray in the same manner, he shall say in the same manner [that] they are to go(?).
Then the prinuati shall go back(?) by the same way by which you will have come.’ Com-
pare VIb 56f., describing the end of the first circumambulation: Ape ambrefurent term-
nome benurent ‘when they will have circumambulated fand] will have come to the termi-
nus’, with both verbs in third plural active.

I have here deliberately rendered couortuso and benuso with the ambiguous English you,
leaving it unspecified whether the verbs are singular or plural. Nussbaum takes them as
singular, referring to the individual to whom the instructions for the lustration ritual in its
original formulation were addressed generally. (Cf. Ib 10 pune puplum aferum heries = Vib
48 pone poplo afero heries ‘when thou wilt want to circumambulate the people’.) That is

certainly possible. But in each case at least three people are doing the moving — “he who
will have the perca arsmatia”, and the two prinuati (prinuatur dur Vib 50). Hence the
plural verbs ambretuto, pesnimumo, and the termnome benurent of VIb 57 corresponding
in the third person to the zermnome couortuso and termnome benuso of VIb 63f. and 64.

Pragmatically, therefore, it is at least equally easy to take couortuso and benuso as plurals,
referring to the entire group making the lustration, and, at least in the original formulation,
agreeing in person and number with the plural imperatives pesnimumo:

63f. Ape termnome couortuso, sururont pesnimumo
‘when ye will have returned to the terminus, ye shall pray in the same manner’;

64f. Ape termnome benuso, sururont pesnimumo
‘when ye will have come to the terminus, ye shall pray in the same manner’.

In 65, prinuatur simo etuto erafont uia pora benuso
*the prinuati shall go back(?) by the same way by which ye have come’

evidently correlates with

Vib 52 uia auiecla esonome etuto

‘they (viz. he who will have the perca arsmatia and the two prinuati) shall go to the cere-
mony by the augural way.’ Contrast the formulation in Ib 20f. Puni amprefu(u)s, persnimu
‘when thou wilt have circumambulated, thou shalt pray’, where the clear singular ampre-
fu(ujs is correlated with the clear singular persnimu.

9. Nussbaum thinks that couortuso and benuso are deponent in form, with the second
singular secondary ending *-so of Greek (e.g. Hom. 5é£o), Iranian (e.g. OP yadaiSa), and
Latin (-re in all tenses and moods, sometimes extended to -ris outside the imperative, rus
in some dialect forms). He supposes that the perfects of *gem- and *wert- were originally
deponent in Umbrian, before being remodeled to the active shapes attested as benus IIb 16,
benust VIb 53, benurent Va25,28, Vb 5 benurent VIb 57 and vurtus Ila 2, kuvurtus Ib 11,
couortus Vila 39, courtust Vla 6. When Tables VI and VII were inscribed he thinks that
couortuso, benuso were not replaced by the new active forms either because the engraver
did not understand them and so left them unchanged, or because he interpreted them as
impersonals with an ending -ofr?).

Phonologically, there is no difficulty with this view. The objection in section 5 to Wacker-
nagel’s derivation of the imperative plural -tu, -ta, -to from *-to does not apply here, since
there are no forms in the native alphabet of the type $benusa, $Kuvurtusa.

But morphologically I think the difficulties are considerable. Non-periphrastic deponent in-
flection is quite unknown elsewhere in the Osco-Umbrian perfect. In Latin, where we know
the full system, we can say that the non-periphrastic perfect was activum tantum (on the
active origin of the first sg. -i cf. e.g. Szemerényi (1970: 226), and on the third pl. -ere
from *-eri see Peters (1981: 102)). Indeed, Latin even has a few active perfects correlated
with deponent presents, viz. reverti : revertor, memini : reminiscor (Leumann 1977: 509).
Latin deponent perfects are formed with copula plus /o-participle, just like passive perfects,
e.g. nactus sum, ausus sum like captus sum. The reason for the lack of non-periphrastic
mediopassive perfects in Latin is presumably that the Proto-Indo-European perfect lacked
a distinction of voice, and Latin never created a distinct mediopassive perfect (unlike Indo-
Iranian, Greek, and Celtic), instead conforming its preterits of aorist origin to its perfects

by abandoning mediopassive inflection in such stems. (But not in the type iussitur, mercas-
situr, which was not incorporated into the perfect system.)

Aside from couortuso, benuso the extant Umbrian material conforms to Latin, with ex-
clusively periphrastic mediopassives within the perfect system, whether passive like ukar
pihaz fust Ib 7 = ocar pihos fust VIb 46f. ‘the citadel will have been ritually cleansed”, or
deponent like pue persnis fust VIb 39 ‘who will have prayed'. (Cf. Rix (1976a) on the
correct analysis of forms in -//e/i as present passive infinitives.) Oscan does seem to have at
least one non-periphrastic passive within the perfect system, comparascuster “will have been
discussed’ Tab. Bant. 4. (I omit discussion of such problematic forms as /amatir Tab. Bant.
21, lamatir Vetter 6.4, sakrafir Vetter 86.9f., 87.9.) But this is passive in value, not de-
ponent, and is an easily understandable innovation (as if Latin were to form $ceperitur
beside captus erit); and Oscan also has periphrastic passive perfects and pluperfects, e.g.
ligat[us] fufans ‘had been delegated’ Cip. Ab. A9f., pruftuset ‘have been approved’ ibid.
A 16, scriftas set ‘have been written’ Tab. Bant. 25. (But ancensto fust ibid. 22 is probably
adjective plus copula ‘will be (in) unassessed (state)’, not periphrastic future perfect ‘will
not have been assessed’.)

It is therefore not easy to suppose that Umbrian once had deponently inflected future
perfects to the active presents attested in menes ‘thou wilt come’ Ib 15 and kuvertu Ib 9
etc. couertu Vib 47 etc. ‘thou shalt/he shall return’.

10. If we take benuso, couortuso as active plurals, the difficulty of explaining how Umbrian
would have the mediopassive ending *-so in its future perfect disappears. And we can
suppose that the mistake that the redactor of VI and VII made was simply that, misled by
the ambiguous neighboring ambretuto, pesnimumo, etuto, he failed to change second
plural couortuso, benuso into third plural *couorturent, benurent (as he remembered to do
at S6f. ambrefurent . . . benurent), i.e. exactly the same mistake that Nussbaum has so con-
vincingly demonstrated for the singulars in -us instead of -ust in VI and VII.

But a new difficulty arises. Although the origin of the Osco-Umbrian future perfect formant
-us- is obscure, it at least combines regularly with the personal endings second singular -s,
third singular -t, third plural -ent to give Umbrian -us, -ust, -urent; cf. Oscan third singular
dicust etc., third plural tríbarakattuset, angetuzet, pass. comparascuster. I would therefore
expect -us- to combine with the second plural ending -fo < *-ta to give §-usto, not -uso.
Oscan and Umbrian have numerous cases of st which apparently go back to Proto-Indo-
European *st, e.g. Umbrian pustnaiaf Ib 22, and it does not seem feasible to explain all of
these by analogy or special sound laws.

However, the difficulty of an ending with -s- for expected -st- in Osco-Umbrian is not uni-
que to Umbrian couortuso, benuso. In the Paelignian inscription Vetter 213, Thurneysen
long ago (1888: 352) proposed to take /exe in line 7 as the equivalent of Lat. legistis, i.e.
a second plural perfect in which the ending -fe had somehow combined with the Paelignian
equivalent of Lat. -is- to give -se. Thurneysen thought of simplification of the consonant
cluster xt of *lexte syncopated from */égiste. But von Planta (1897: 283) pointed out that
we might expect “lekste to be simplified rather to §leste.

It looks instead as if, for reasons that now escape us (partly because our knowledge of the
languages is so fragmentary), the -/- of the second plural endings -to and -te combined in
Osco-Umbrian with a preceding -s- to give an -s- that was, apparently, short (at least in

Umbrian), but differed from inherited *s in not being subject to rhotacism in Umbrian.
How this came about, I cannot say, but Paelignian /exe and Umbrian benuso, couortuso
support each other. As often in historical linguistics, to the theory and methodology of
which Henry Hoenigswald has contributed so much, two parallel cases, each of which by
itself would be rather dubious, gain greatly in verisimilitude by the very fact of their being
two. (As von Planta (1897: 366) points out, is is not clear whether lexe is a suffixless
preterit like Lat. /egi or an s-preterit like Lat. intellexi; in the former case, it is evidence for
a stem extension -s- or -Vs- in the second plural perfect in Osco-Umbrian parallel to the -is-
of Lat. leg-is-tis.)

11. It appears therefore that in the distribution of the Proto-Indo-European second person
endings plural *-te and dual “tes, *-ta the different subbranches of Italic and Celtic have
gone very different ways. Latin and Irish seem to have retained plural *-te as a distinctive
imperative ending, while using dual *-tes for the other moods. Paelignian, on the evidence
of imperative eite and indicative /exe, generalized plural *-te in all moods, while Umbrian
generalized dual *-ta in all moods. For Oscan proper and Continental Celtic we lack evi-
dence," and the British languages have undergone too much change for us to say which of
these three endings their second plurals continue.

The shapes and distribution of the second and third person dual endings in Proto-Indo-
European remain unclear. Umbrian adds a witness to Balto-Slavic for a second dual in
*.t4, but the original distribution and antiquity of this vis-à-vis the *-tes of Irish, Latin,
Gothic, Slavic, and Indo-Iranian remains problematic, as well as its relation to the secon-
dary third dual *-tam of Sanskrit and Greek, and of all these to the *-tom of Greek -Tov,
Indo-Iranian -tam, Toch. B. -tem.$ (Quite unclear the Tocharian preterit third person duals
B stamais, ltais, A takenas, except that, like all Indo-European active second and third
person duals, these can be described as second plural endings with additional marking: in
this case, the second plural endings are those of the Tocharian preterit and imperative, type
B rankas, pkarsas, A kotas, ptärkäs.)

A final note on the non-imperative second plural mediopassive of Umbrian. If the impera-
tive -mo, -mumo has been (re-)shaped on the model of active imperative -to, -tuto (as if
Lat. impv. -mini had been replaced by §-mine on the analogy of active imperative -te),
and if the active has -to in indicative benuso, couortuso as well as in imperative, then there
is every reason to suppose that the second plural indicative and subjunctive mediopassive
ending of Umbrian also remodeled its original shape (*-mu(s], *-mur from nominative
plural masculine *-n(nJos?) to *-mo (as if Latin indicative/subjunctive -mini has been re-
modeled to *-minis after active -tis).

Notes

! Prosdocimi (1979: 212) proposes to restore Jmod in an early inscription from the neighborhood of

Tivoli, with an ending analogous to Umbrian -mu, but in a Latin dialect. While the final -d is just
what we expect in Early Latin, -m- instead of -m(VJn- in a Latin dialect would be very surprising,
and I wonder if it is possible to read the traces of the first visible letter as zt, giving an imperative in
-m(V]]nod. On the likelihood that Luwian -mmi- and Balto-Slavic *-ma reflect *-mno- see H.C. Mel-
chert (Die Sprache 29. 1983: 23-25).

? ] am not sure enough about the meanings of eta-, arsma-, and catera- to venture translations for these

imperatives. Eta- is generally taken as equivalent to Latin ita- ‘be in the habit of going’, but a fre-
quentative seems out of place here, nor is it clear to me that the meaning ‘go’ fits. The command
etato (subj. etaia(n)s in indirect discourse VIb 64, 65 (=VIIa1)) is addressed to the men of Iguvium
after the prayers following each of the three circumambulations of the lustration. It cannot well
mean ‘go (away, because the ceremony is over)’ — that would be appropriate only at the end of the
part of the rite that involved the men of Iguvium generally. It is scarcely likely that the men of
Iguvium are to go away three times during the ceremony, presumably returning at least twice, and
nothing is said about a return of the men of Iguvium after these putative departures. Nor, as far as I
understand the text, do the men of Iguvium as a whole take part in the circumambulation; rather
they remain in the same spot, and ‘he who has the perca arsmatia’ and the two prinuati do the
circumambulating. And the command efato/etaians is given after each of the three circumambula-
tions, not before.

Lejeune (1949: 107) has shown that this vowel was probably shortened in Umbrian, as it evidently
was also in Oscan (Lejeune 1949 passim).

I leave undecided whether this is an old variant *-£e, or represents analogic remodeling after forms
which retained inherited long vowels and diphthongs before enclitics and shortened them in absolute
final.

Léon Fleuriot has suggested (EC 17. 1980: 135), that Gaulish Joncate is a second plural indicative
‘vous avalez’; and one year later (Fleuriot, EC 18: 1981, 91-92) he has suggested that Gaulish ibetis
and biiete are second plural imperatives, ‘buvez’, ‘soyez’, the former with an enclitic pronoun -s. If
these interpretations are correct, Gaulish, like Paelignian and Greek, but unlike Latin, Irish, and
Umbrian, preserved Proto-Indo-European plural *-te in both indicative and imperative, without
intrusion of dual forms.

The -m here is of course not the *-m of the Proto-Indo-European ending, which after *o wouid have
been lost without trace in Tocharian, cf. e.g. B ce ‘hunc’. It is presumably the same third person
pronoun as in the B third singular present nesäm etc.

References

Brugmann, Karl, 1892. Grundriß der Vergleichenden Grammatik der indogermanischen Sprachen 2/2
` (Straßburg: Trübner).

Brugmann, Karl, 1911/12. “Zu den Imperativendungen im Umbrischen”, Indogermanische Forschungen
29: 243—249.

Brugmann, Karl, 1916. Grundriß der Vergleichenden Grammatik der indogermanischen Sprachen 2/3
(2nd edition) (Straßburg: Trübner).

Buck, Carl Darling, 1928. A Grammar of Oscan and Umbrian (2nd edition).

Bücheler, Franciscus, 1883. Umbrica

Bugge, Sophus, 1855. *'Die Formen der geschlechtslosen persönlichen Pronomina in den germanischen
Sprachen”, Zeitschrift für vergleichende Sprach forschung 4: 241—256.

Cowgill, Warren, 1975. “The Origins of the Insular Celtic Conjunct and Absolute Verbal Endings”, in:
Flexion und Wortbildung (= Akten der V. Fachtagung der Indogermanischen Gesellschaft) edited by
H. Rix (Wiesbaden: Reichert), 40—70.

Cowgill, Warren, 1979. "Anatolian Ai-Conjugation and Indo-European Perfect: Instalment II”, in:
Hethitisch und Indogermanisch (= Innsbrucker Beitrige zur Sprachwissenschaft, Band 25) edited by
E. Neu and W. Meid (Innsbruck: Institut für vergleichende Sprachwissenschaft), 25-39.

Einarsson, Stefan, 1945. Icelandic Grammar, Texts, Glossary (Baltimore: John Hopkins Press).

Kronasser, Heinz, 1965. Etymologie der hethitischen Sprache 4 (Wiesbaden: Harrassowitz).

Lejeune, Michel, 1949. “Sur le traitement osque de *-a final”, BSL 45: 104-110.

Leumann, Manu, 1977. Lateinische Laut- und Formenlehre (2nd edition) (München: Beck).

Nussbaum, Alan, 1973. “Benuso, Couortuso, and the Archetype of Tab. Ig. I and VI-VIIa”, The Jour-
nal of Indo-European Studies 1: 356--369.

Peters, Martin, 1981. “Notice of Lapis Satricanus. Archaeological, epigraphical, linguistic and historical
aspects of the new inscription from Satricum””, Die Sprache 27: 101f.

Planta, Robert, 1892. Grammatik der oskisch-umbrischen Dialekte 1 (Straßburg: Trübner).

Planta, Robert, 1897. Grammatik der oskisch-umbrischen Dialekte 2 (Straßburg: Trübner).

Prosdocimi, A. L. 1979. “Studi sul latino arcaico”, Studi etruschi 47 (serie 3): 173-221 (with Tavole
43-51).

Rix, Helmut, 1976a. “Die umbrischen Infinitive auf -fi und die urindogermanische Infinitivendung
-dhidi”, in: Studies in Greek, Italic and Indo-European Linguistics offered to L. R. Palmer, edited by
A. Morpurgo Davies and W. Meid (Innsbruck: Institut für vergleichende Sprachwissenschaft) 319—
331.

Rix, Helmut, 1976b. “Subjonctif et infinitif dans les complétives de l’ombrien”., BSL 71: 221—240.

Rix, Helmut, 1976c. "Umbrisch ene ... kupifiaia", Münchener Studien zur Sprachwissenschaft 34:
151-164.

Szemerényi, Oswald, 1970. Einführung in die vergleichende Sprachwissenschaft (Darmstadt: Wissen-
schaftliche Buchgesellschaft).

Thurneysen, Rudolf, 1888. “Zur Inschrift von Corfinium", Rheinisches Museum für Philologie, Neue
Folge 43: 347 —354.

Vaillant, André, 1966 . Grammaire comparée des langues slaves. Tome 3/1: le verbe (Lyon: IAC).

Vetter, Emil, 1953. Handbuch der italischen Dialekte 1 (Heidelberg: Winter).

Wackernagel, Jakob, 1907. “Indisches und Italisches. 5. Umbrisch efato”, Zeitschrift für vergleichende
Sprachforschung 41: 318—319 (= Kleine Schriften 507-508).

Wackernagel, Jakob, 1912/1913. “Lateinisch-Griechisches”, Indogermanische Forschungen 31: 251—
271 (= Kleine Schriften 1228—1248).

heres, xnpworat: indogermanische Richtersprache

George Dunkel (Princeton University)

“Alter familienrechtlicher Ausdruck,
der Bildung nach mit unorns ver-
gleichbar” H. Frisk

1. Pott: *ghed-

In 1867, A.F. Pott (1867: 68—71) compared Greek xrjpa ‘widow’ with Latin heres ‘heir’,
and ‘sniffed out’ (witterte) in the latter an old compound, with a verbal second element.!

As the first member, Pott saw a *-ro- derivative of the root *gheE- ‘abandon’ (Pokorny
1959: 418), as in xjpa and the later xnpos ‘widower; bereaved, empty’. This much of his
analysis has remained generally accepted, although *-ro-derivatives generally show the zero-
grade of the root (Brugmann 1906: 34850).

The identity of the second element has been more controversial. Pott considered three
possibilities: the root *Eed- ‘eat’ (Pokorny 1959: 287ff);? an Indo-European form related
to Indic @ da- ‘take’; and the root *ghed- of Lat. praehendo, Engl. get, etc. (Pokorny 1959:
437—438). Pott thought the last alternative to be the simplest and most likely; his deriva-
tion, though not stated as such, was *ghero-ghed- > *hero-hed- > hered-.

2. Prellwitz: *Eed-

But this was rendered impossible by Prellwitz’s adduction of the archaic Greek legal term
xnpworat. Assuming in the latter a secondary recharacterisation by the agent-suffix *-ta-,
proto-Greek *kherod- and Latin heréd- together permitted the reconstruction of an
ablauting, athematic *ghere/od- Prellwitz saw in the second syllable the verbal root *Eed-
‘eat’, but as support offered in the first place citations from Horace.’ This seems to have
convinced no one, with the single exception of M. Leumann (1977: 391, 393).

LSJ define ynpworar as ‘far-off kinsmen, who seize and divide among themselves the prop-
erty of one who dies without heirs (xnpos)’. The two early hexameter attestations, namely
Il. 5.155—8 &v0' Ó ye Tovs Evdpite, ptrov 5' é£atvvro Bvuòv

auporépuw, TaTEpı 66 yoov Kat Kndea AVYPAa

dein’, ered ob CWovTE udyxns ‘EK VOOTHGAVTE

è étaro: xnpworat dé Sua KTmRow Saréovro.

and, with enjambement,

Theog. 604—7 ...O0Ào0r 6 €nt yrpas iknrat
XNTEL ynpoKopuow | ó 600 Buórov Vem euns
twe, anopdıuevov Bè Sia £corv 6aréovrai
xnpwoTal'...

are paradigms of the dangers of lack of sons, a well-known worry of e.g. the Vedic Aryans
as well. In both passages, a dearth of direct descendents, due to slaughter or celibacy, leads
to a man's property being divided up (61a Saréouar) by the xnoworat.* Although it is not
unlikely that the xnpworai were distant relatives, nothing in the texts supports this view at
all; the paradigms become all the more horrible if outsiders, rather than kinsmen, divide up
the ‘abandoned’ inheritance.

3. Brugmann: á då-

Recent scholarship? has generally followed Brugmann's resolution of Pott's triple choice
(Brugmann 1904: 30; 1906: 79, 396f.). For him *Ked- as second element, in a compound
meaning ‘das Ererbte verzehrend', was as impossible for semantic reasons as was *ghed-
phonologically, since, as Pott had put it, such a metaphor “‘hiesse dem ernsten, und zumal
in Rechtssachen keinen Spass verstehenden Römer einen übel angebrachten Witz unterle-
gen” (Pott 1867: 69, cited approvingly by Brugmann 1904: 30).

Brugmann was thus left with Pott’s third etymological possibility: connection with Indic
d dd- “take’.° The original meaning would thus have been Erbempfänger’, as in Skr. dayada-
(Manut), cited already by Pott. And a European equivalent to the Indo-Iranian preverb *4
would have been ascertained.

Although I would share Brugmann's joy at finding a clear European counterpart to I—Ir.
*q, I doubt that *ghere/öd- contains anything of the sort. Despite the near-universal ac-
ceptance of Brugmann’s analysis, various problems remain unresolved. Formally, what
happened to the final consonant of the root, i.e. the *-O in *deO-? Brugmann spoke of loss
before vocalic endings in the paradigm, but this means taking Lat. hered- as only secondarily
athematic in the nominative — rather unlikely in view of Gk. *kherod-. Brugmann further
compared the Indo-Iranian ‘infra-zero-grades’, as in 1st pl. pres. act. da-d-más, vbl. adj.
d-t-ta-, cpd. devd-t-ta-. Greek *-od-tas from *-o-0-dO-teA- would be the only instance in
Greek of such syllabic loss.” But despite the constant progress of laryngeal studies, this is
still more of an ‘obscurum per obscurius’ than an explanation.®

The preverb also poses problems. Frisk and Beekes (1975: 9f.) have rightly objected to the
supposed presence of both ablaut forms, *é and *0, in the same word. Furthermore, the
sense ‘take’ of Indic d da- is due in my opinion to the diathesis alone (exclusively middle in
the Rigveda), not to the preverb.? Other semantic problems are noted in section 11.

4. Beekes: suffix *-ed-

Beekes, who rejected *Eed- as second member “on account of the meaning” and an à dé.
equivalent for the apophonic reasons mentioned above, has recently suggested a return to
another old belief: that the second syllable was not a root at all, but a suffix (Beekes 1975:
9f.).!° Starting from an original nom. *@héEr-dd(-5), acc. *-éd-m, gen. *-d-ós,! varying
generalisations would account for the disparities between the Greek and Latin vocalisms.

This proposal, cited approvingly by Chantraine, is indeed a plausible resolution of the
diverging vowels — but of little else.!? First, the comparative evidence for a suffix *-ed-
is extremely limited: the only other instance provided!” was I—Ir. *sarad-/sard- ‘autumn’,
itself very obscure.!*

Further, Beekes made no mention of the value of the suffix in his reconstructed *gheEr-0d-.
Is it possessive, agentive, or something else? Our delight in powerful new paradigmatic
models does not free us from the need to explain the function of the reconstructed mor-
phemes.P?

Finally, a segmentation *gheEr-od- separates the first element from the thematic *ro-
derivatives of Greek; this too needs explanation.

In any case, we could still accept Beekes’s ablaut, if we wanted,' without feeling obliged
to avoid the semantic question as well.

5. My view is that only the derivation from *Ked- ‘eat’ is correct. I believe in the utter
exactitude of both clauses of Frisk’s initial sentence quoted at the outset." Proof comes
from the texts, rather than from considerations of ablaut or paradigms. Still, let us note at
the start that Frisk's formal objection to this interpretation, "gegen Anknüpfung an ed-
spricht... der w-Vokal”, is groundless. The root *Eed- is known to occur in both long-
and full-grades (Narten (1968: 15, n. 44); differently Watkins (1969: 31—32)), both e- and
o-colored (Schindler (1975: 61ff.)). Regardless of which explanation of the variant forms
appeals to us most, allomorphy will not prevent us a priori from finding a form of *Eed-
in our word.

My suggestion is to begin from *gheEro-Eed-. The dialectally varying contraction-colors are
comparable to those in the thematic ablative singular: *-o/A jad gave *.6d in Latin, Greek,
and Celtiberian (see Untermann (1967: 281ff.)), but *-ad in Baltic.

A pre-form *gheEro-Eed- is morphologically plausible because, in contrast to the isolated
and meaningless "suffix *-ed-", a small cluster of parallel verbal government-compounds
with *Eed- as second member is traceable. Most obvious is *Omo-Eed-, as preserved in RV.
ámád-, Gk. couno-rns'” (with the same recharacterisation in *-td- as in xnoworts);
perhaps old is also *medhu-Eed-, as in RV madhvdd-* and, with long-grade and substan-
tivised, OCS medv&dt ‘äpkros’. All but the last are found as epithets of birds?’ and carrion-
eaters.

6. Related phraseology: Greek

The Odyssey describes an attempt at usurpation of an inheritance by non-kinsmen who
ignore a legitimate heir’s right to his property, and, what is more, plot to kill him. The
suitors’ activity is metaphorically termed ‘eating the house’. An undertone of disapproval
and blame accompanies this violent image:

1.250-1 ... Tor dé PE wovaw ébovrec
oikov Euov- Taxa 81 ue dappawvovor Kai abrov.
2.237-8 opàs yàp Tap0éuevor KepaXàs karéSovot fiaiws
okov Odvoonos...
4.318 'eodietai dou oikos, OAWXE dé Tova Epya:

16.431 Tou vuv otkov armor bens...

21.332--3 oË SÙ otkov arıuafovres EB ovow
avi pos aptoTnos....

Feast imagery, referring to the suitors’ gluttony, is of course widespread all through the

Odyssey,” and at times is broadened to include consumption of ‘the wealth’ in general:

11.116 dvi pas ùmeppiarovcs, ol’ ToL Biorov Karedovoı

13.419 ...Biorov dE ol AAAoı Eöovon.

14.377 nö ot xatpovow (orov viyrotwov é60vrec -

17.378 ...Bworov karé&ovow üvak Toc

and, with other direct objects,”

16.389 un oi xpnuar ... é6couev

23.9 ..kTrrjuaT é60v BuocovrÓ re maí6a.

With the suppletive aorist stem $ay-:

3.315--6 Mt ToL Kara Tavra daywoL
KTUATA ÜaATOAUEVOL..

16.428--9 Tov p €Gedov plivar Kar atmoppaica: dixov Dron

NSE Kata Swnv dayéew pevoetk éa TOAAHV `
In this wider context, we might wonder how much of the feast imagery is the Odyssey-
poet’s own contribution. My opinion is that since only the expression ‘eating the house’ is
found as well in related traditions in the same meaning (dishonorable, potentially violent
appropriation of property), only it is inherited; the rest we can ascribe, for the moment
at least, to the elaborating hand of the monumental Odyssey-poet.??

7. Related phraseology: Hittite

The Odyssean house-eating expression is paralleled in one of the oldest Hittite texts: the
proclamation of Telipinu (CTH 19). This attempt at regulation of royal succession begins
with a detailed catalog of internecine murders and usurpations. I. 21—2 is a passage which
has long caused problems for Hittitologists:

man appizziyan=ma iR. MES DUMU.MES LUGAL mar¥e¥8ir
nu E.MES-SUNU karipuwan dair

išhašaš=ma=šan taštašeškiuwan dair

nu ešhar=šumit eššuwan tier

“Als aber später die Knechte der Prinzen betrügerisch wurden, begannen sie, ihre Häuser
zu fressen, gegen ihre Herren immer wieder Verrat zu üben und (immer wieder) ihr Blut zu
vergiessen.” (Eisele (1970: 21)).

The most recent editor, Eisele, calls the passage “nicht ganz klar”, and offers two possible
interpretations: either (following Friedrich) the servants destroyed the princes’ houses, or
(following an oral suggestion of Sommer), “auffressen = Verwirtschaften der Güter, die die
Knechte für die Prinzen bewirtschafteten, zu eigenen Gunsten." (Eisele (1970: 73); I quote
here from Kammenhuber (MIO 3. 1955: 34ff.)). Similarly Hoffmann (1984: 17).

But the similarity with the Odyssean metaphor is unmistakable, despite the lexical re-
placements (pir/parn- (2 É) for *dom- or *wik- continuants; animalistic, 'daevic! karep-?
rathet than ed-). Corresponding to the suitors we have the 'princes' servants! who begin to
‘devour’ their masters’ ‘houses’. The sense must be that the servants began to intrigue so as
to usurp royal property and thereby royal power.”®

8. Related phraseology: Rigveda

The Rigveda too is full of references to division of wealth, though usually in the form of
prayers for a share in the booty taken from a defeated pagan enemy. In one stanza, how-
ever, eating terminology must refer to family property. These mantras occur in a series of
riddle-verses addressed to Indra:

2.13.4abc prajabhyah pustim vibhdjanta dsate
rayim iva prsthdm prabhavantam ayaté
asinvan ddmstraih pitur atti bhojanam

‘They sit, distributing wealth to the offspring -- riches like a peak, prominent to one who
approaches. dsinvant- with (his) teeth, he eats the wealth of the father.”??

Both pusti and bhójana- can mean ‘food’ as well as ‘wealth’ in the Rigveda. And food and
wealth are mixed up in the complex simile in 5.?? But formulaic analysis shows that both
must be taken in a non-alimentary sense here: for the first, cf. our initial alliterative
formula in another wealth context:?!

8.59.7bc rayás pósam ydjamanesu dhattam
prajam pustim bhütim asmasu dhattam

For the second, cf. the common formula which is split between verses a and c in our
passage:

2.26.1d ydjvéd áyajyor vi bhajati bhöjanam

10.48.1d ahäm däsuse vi bhajami bhôjanam

In fact, vi bhaj- is the term normally used of disposition of the wealth of the pagan enemy,
as in

10.84.2c hatvdya sátrün vi bhajasva védah
10.27.10d áyuddho asya ví bhajani védah
1.103.6d áyajvano vibhájann éti védah

73270 vi tvahatasya vedanam bhajemahi, etc.”

Functionally, vi bhaj- corresponds to 61a Saréopa, the characteristic activity of xnpworai.
Further, öı4 and vi are etymologically related (KZ 95 (1981): 230), while bhaj- is etymo-
logically cognate with the Greek stem ¢ay-, forming an aorist to Soft, The language
seems old.

The wealth being eaten is characterised as pitur, ‘of the father’. Here I must reject Ber-
gaigne’s artificial mythologic explanation, accepted by Oldenberg,?? in favor of the simple
interpretation that a father is the recent decedent, cf.

1.70.10d pitür nd jivrer vi vedo bharanta

In other words, we are dealing here not with the booty of pagan enemies, but with a famil-
ial disposition. The unexpressed subject of asate in a is something like ‘the family, the
clan’.

The meaning of dsinvant- is unclear,** although given the riddle context, Geldner’s ‘‘ohne
Bissen zu machen ißt er mit den Beißzähnen” sounds plausible.”

The general picture is thus that of a ceremonial distribution of a dead elder’s wealth among
his offspring. The high pile beckons, however, also to the approaching potential usurper,
who can ‘eat’ without chewing. We recognise a now-familiar use of ad-.

9. Evaluation of related phraseology

In short: early Greek, Hittite, and Vedic texts speak of ‘eating’ or ‘devouring’ property in
reference to the usurpation of a household by others than a direct son of the ‘pater famil-
ias’. This seems indeed to be an “alter familienrechtlicher Ausdruck”.

This idiomatic, metaphoric phraseology shows that the verbal root *Bed- is in fact a
perfectly plausible second member of "éheEro-Eed-. The original meaning could only have
been ‘who eats what has been abandoned', i.e. the house and property of the departed
father.” Rather than causing semantic problems, recognition of *Eed- 'eat' in this com-
pound fits well into the violent, extra- or marginally-legal semantic sphere of the usurpative
phraseology.

10. Animal imagery for anti-social acts

More support for this interpretation comes from a more general consideration of animal
imagery referring to disruptive behavior in early dialectal texts. As is known, the direct
naming of dangerous or annoying animals (wolf, bear, ant, wasp) was avoided in polite
Indo-European conversation, on the nomen est omen principle. This led to circumventive
strategies such as taboo deformation, replacement by an epithet, and other periphrases.

In extreme instances of asocial behavior, however, the breaking of this taboo seems to have
been societally sanctioned. E.g. in wartime, an accepted feature of heroic conduct was
insult before combat. This could include the mocking, proleptic description of the feast
that vultures, fish, dogs, worms, or other flesh-eaters would make of the opponent's body
as it lay on the ground. For the moment,? cf.

RV 10.95.14cd ddha säyita nirrter upasthé
‘dhainam vrka rabhasaso adyuh

‘There would he lie in the lap of death, there would ravenous wolves eat him’, with

Il. 22.42—3 ..Táxa. kev € KÜvec kat yórec ébouwv[ketievov...

In particular, the taboo on the invocation of the wolf seems to have been suspended to
mark rejection from the clan after destructive behavior or crime. E. Campanile has shown
that “in the juridic language, ‘wolf’ was the technical denomination of the man banished
from his people" (JIES 7. 1979: 237ff.) by comparing phraseology from Hittite,?? Indic,??
Germanic, and Celtic. Cf. also Livy 1.4.7 sunt qui Laurentiam vulgato corpore lupam inter
pastores vocatam putent. The reason for the suspension of a taboo so strong that it seems
to have led to metatheses and dissimilations already in Indo-European can only have been
the extreme usefulness of the animal imagery in referring to censurable behavior.

Now *omo-Eed- and *medhu-Eed-, we remember, occurred as epithets of predatory birds
and other animals (section 5). So too, the continuants of *éheEro-Eed- and related ex-
pressions were also closely associated with gore, as in Odyssean taxa 57 ue Suappatoovot
Kai abróv, opas yàp mapBéuevor xedaXàs karésovot Bawsloikov, pIlvai Kat atoppaica,
etc.: and Telipinu nu eXhar-Xumit eXsuwan tier; karep- ‘fressen’ for ed- ‘essen’, etc.

In short: animal imagery was used to refer disapprovingly to disruptive, disorderly behavior,
and the *-Eed-compounds formed part of this animal imagery.

11. Connotations of *gheEro-Eed-

Although Latin heres indicates the heir in general, xnpcoorat are something rather more
specific: appropriators of a vacant inheritance. Benveniste has suggested that Indo-Euro-
pean possessed no general word for heir at all. Such a term was not needed in a patriarchal
society, where property normally passed only from father to son. But a term might have
been useful for deviations from this practice.^!

*gheEro-Eed- seems to have been an accepted metaphor of the indogermanische Richter-
sprache for collateral heirs and usurpers of households outside the direct line of descent.
Brugmann and Benveniste, with their bland ‘Erbempfanger’ and ‘héritier’, back-project a
later, more abstract legal system into Indo-European times. They ignore as well the animal-
istic semantic component suggested by related phraseology and by the form’s composi-
tional cluster-mates. Far from jocular, the connotation of “gheEro-Eed- seems to have
been that of a near-taboo, asocial scavenger of “abandoned” property. Time and social
change have caused semantic fading: to "collateral heir", then to "heir", and even to
"relative" in general, if Hesychian oi uakpo0 ev ovyyeveis is to be taken seriously.

The presence of metaphor, a feature of the Dichtersprache, in the Richtersprache as well
is not surprising. Formulaic expressions, no-longer-understood archaisms beside up-to-
date expressions, parallelisms of structure, and other seemingly oral poetic features are
common in legal texts, and the relations between law and poetry have been repeatedly
explored. Metaphor is in any case a constantly self-renewing feature of human language
in general: though highly developed in, it is by no means limited to a poetic style.

Notes

Ir thank the Alexander von Humboldt-Stiftung, my university, and my department for the stays in

Würzburg (1980-1981) and Paris (1982-1983) during which this paper came into being.

On the laryngeal see Schindler (KZ 89. 1975: 61ff.); but the argument of this paper is in no way
dependent on the laryngeal.

Cf. Prellwitz (BB 25. 1899: 313--314). He did however also mention the compound-cluster discussed
in section 5.

For normal distribution among living direct sons, cf. Od. 14.208—9 ...roc 8€ t«v t6ácavro | rai6eq
bnépOvuo. kac ent KAnpovs eBdAovTo. The Saréoua phrases all fill the line from the masculine
caesura onwards.

Boisacq (1938), Walde-Hofmann (1938-1956), Schwyzer (1939--1950), Pokorny (1959-1969),
and Frisk (1960-1972) s.v.; Ernout-Meillet (1951-1952) reject summarily all explanations of the
second element, without proposing one of their own.

Which he considered to have been the ultimate source of his *ed- ‘eat’ anyway (Brugmann 1904:
31)!

Nothing comparable is mentioned by Szemerényi (1964).

None of the explanations of the Indic infra-zero-grades (e.g. generalisation from 3rd pl. *de-dO-nti
(Meillet 1922: 66); automatic non-distinctive overlength-avoiding syllabicity (Hoenigswald 1965:
97-99)) will work for xnpworns.

More on the preverb elsewhere; for now, see (ZIJ 24. 1982: 89ff.).

Older believers in a suffix *-ed- are listed by Brugmann (1904: 30).

A paradigm he called hysterodynamic and others would call holokinetic.

Beekes is followed, with reservations, by Hamp (KZ 96. 1982--1983: 101). But Hamp does not
address the weaknesses raised here.

10
il
12

15
16
17
18

RE B

BRE

More material in Brugmann (1906: 468f.); Wackernagel-Debrunner (1954: 174—175). Within Latin,
on cuppes 'gluttonous' see Leumann (1977: 367, 372); on custos, see H. Nowicki (KZ 92. 1978)
(convincingly < *kudh-to-sd- ‘der bei etwas Verborgenem sitzt’); merces ‘payment remains com-
pletely obscure. `
Roots suggested include 1 *kel- ‘warm; cold’ (Pokorny (1959: 551): Lat. caleo, calidus etc.) or 2
*Ker(E)- ‘grow’ (Pokorny (1959: 577): Lat. Ceres, creare etc.).
Again, what is the value of the suffix in *sarad-?
There is no evidence whatever for a weak stem *gheEr-d-.
Which Frisk disregarded in the rest of his discussion.
Only 10.87.7d (hate-song) amadah ksvinkas tam adantv énth ‘Let the raw-eating, éta- ksvinka-birds
eat him!” (with figura etymologica).
Wackernagel (1955: 927). The different color of the contraction is due to the different preceding
vowel: *o... 0 avoided due to labial dissimilation.
Only 1.164.22ab, of suparnd- birds in the tree riddle; Thieme (1949: 65). Johnson (JAOS 96. 1976:
248ff.) brings in the "speculative symposium".
Cf. also the group of Latin ornithonyms formed after ficed-ula, Leumann (1977: 312). Thieme
(1949: 65) argues that RV madhvad- means ‘fig-eater’.
wynoriys refers to oiwvor ll. 11.454; to Kuves IL 22.67; to txaues Il. 24.82; to avp (metaphorically,
of Achilles) IZ. 24.207; to "Exté vav bow Theog. 300; to Képfepov xiva Theog. 311.
Normal, non-metaphoric eating and drinking e.g. 14.81; 15.373, 378; 21.69; etc.
With Burov, cf. Swhv in Theog. 606, quoted section 2; with xrAuara, cf. xrfjow in ZI. 5.158 (ibid.).
Other verbs are used metaphorically of wealth, but not of houses, e.g. Sapdanrw, Kkeipw. These too
may be poem-specific inventions.
On futuristic appizziyan, cf. (KZ 96, 1982: 74; 86).
Kronasser (1966: 520); Oettinger (1979: 421). The “Menschenfressertext” (CTH 17.1) repeatedly
uses ed- for cannibalism, not karep-. Cf. also Telipinu II. 73, and Goetze 1928: 79-80.
The expression is not totally isolated in Hittite, but parallels are mostly too fragmentary to be of
help, e.g. the ritual KBo XXIV 93 (CTH 670) II 17-8:

Qo namma ape E.MES Ser katlta...

GIShuimbaza karipanzi. . .

‘and these devour the houses from top to bot[tom, from the ...] and from the huimba-' (7 floor?
roof?). Cf. also line 20 ‘they devour the wall, beams, windows’, again with karipa[nzi.
Verse d is a refrain (= 2d, 3d), yas takrnoh prathamam sasy ukthyah ‘You who did this first, as such
are you praiseworthy’.
prstha- means primarily ‘back, spine’, and secondarily ‘ridge, peak’. Oldenberg considered taking the
simile as referring literally to a sacrifical meal, but ended up with “Gedeihen verteilend wie Reich-
tum, der (jedem) zu der Höhe Herbeikommenden zugute kommt". The translation in the text
assumes "scheinbar vorangestellt" iva in the sense of Oldenberg (1967: 248ff.). Geldner's contorted
reading involves arbitrary ellipsis.
Other passages showing pusti- = ‘wealth’: 2.12.5c; 5.10.3. ; ,
Cf. also 3.30.7b abhaktam cid bhajate gehyam sah, and 10.112.10d abhakte cid a bhaja raye asman.
“Il est définitivement établi qu’ Indra tue en naissant son père Tvastr pour boire le Soma” (Bergaigne
1877: 59).
The standard connection is with *seA- 'satiate’ (Pokorny 1959: 876), thus ‘unersättlich’. Note the
chance resemblance to ofvouat ‘hurt; despoil’, itself usually derived from *twi-n-yo (Pokorny 1959:
1099) rather than from *si-nw-0-. *si- seems not to have been treated in Greek like *su-, on which cf.
Glotta 60. 1982: 58—59.
Geldner’s interpretation rests on Yaska and other native sources. The riddle of c could be thus ex-
pressed in American English: ‘Who can eat the father’s lettuce (= dollars) without using this teeth?”
(Answer: a *gheEro-Eed-).
The semantic development ‘abandoned’ > ‘inheritance’ seen in *gheE-ro- has two sets of parallels.
The first is the oft-cited one of German Erbe: Greek ópoavóc, Latin orbus, both the latter meaning
‘deprived, bereft; orphan’. The other was pointed out to me by G. Pinault: post-Vedic rikthabhay-
‘heir’ (Manu +). Besides bhaj- = gay-, rikthd- (see fn. 41), like réknas., is a term for ‘property, in-
heritance’ derived from *leykW- (Pokorny 1959: 669), semantically parallel to *gheE- ‘leave’.

37 More elsewhere. Purüravas vs. Urvasi is a war of words, not deeds, cf. (JIES 7, 1979: 249ff.).

More on animal imagery in Hittite: E. Neu (1974: 103).

The parenthetic phrase vrko hí sáh at 9.79.3b discussed by Campanile recurs non-parenthetically in
6.51.14d jahi ny atrinam panim, vrko hi sah ‘Strike down the tightwad Atrin, for he isa wolf". Cf.
also 2.23.7ab uta va yo no marcayad anagaso/‘rativa martah sanuko vrkah, and 2.28.10c steno va
y6 dipsati no vrko va. Avestan vahrka- could provide further material, e.g. Y. 9,18.

The xnoworar passage Il. 5.155 —8 also involved merciless slaughter.

Benveniste (1969: 84): “N’étaient dits héritiers que ceux qui héritaient à défaut du fils; c'est le cas
des xnocorac ..." On exclusion of females, cf. Schrader-Nehring (1917—1929 s.v. Erbschaft),
Wackernagel (1920-1924: 28) (heres as fem. is late), and RV 3.31.2a na jamaye tanvo riktham
araik, if the traditional interpretation as ‘the direct son has left no inheritance to his sister’ is correct.
Could our term be a substantivised epithet for an earlier form now totally lost due to taboo?

References

Beekes, Robert S.P., 1975. “Two notes on PIE stems in dentals”, in: Flexion und Wortbildung, edited
by H. Rix (Wiesbaden: Reichert), 9-14.

Benveniste, Emile, 1969. Le vocabulaire des institutions indo-européennes 1 (Paris: Editions de Minuit).

Bergaigne, Abel Henri Joseph, 1877. La Religion vedique d’apres les hymnes du Rig-Veda 3 (Paris: F.
Vieweg).

Boisacq, Emile, 1938. Dictionnaire étymologique de la langue grecque. (3rd edition) (Heidelberg:
Winter).

Brugmann, Karl, 1904. “Altind. dayada-, gr. xnpworns u. lat. heres”, in: Album Kern (Leiden: Brill).

Brugmann, Karl, 1906. Grundriß der vergleichenden Grammatik der indogermanischen Sprachen 2/1
(Straßburg: Trübner).

Eisele, W., 1970. Der Telipinu-Erlaß (Dissertation) (München).

Ernout-Meillet, Antoine, 1951—1952. Dictionnaire étymologique de la langue latine (3rd edition) (Paris).

Frisk, Hjalmar, 1960—1972. Griechisches etymologisches Wörterbuch (Heidelberg: Winter).

Goetze, Albrecht, 1928. Madduwattas (Leipzig: Hinrichs).

Hoenigswald, Henry M., 1965. “Indo-Iranian Evidence”, in: Evidence for Laryngeals, edited by W.
Winter (London/The Hague: Mouton), 93—99.

Hoffmann, Inge, 1984. Der Erlaß Telipinus (Heidelberg: Winter).

Kronasser, Heinz, 1966. Etymologie der hethitischen Sprache (Wiesbaden: Harrassowitz).

Leumann, Manu, 1977. Lateinische Grammatik 1 (2. Auflage) (München: Beck).

Meillet, Antoine, 1922. Les dialectes indoeuropeens (2nd edition) (Paris).

Narten, Johanna, 1968. ‘Zum 'proterodynamischen' Wurzelprüsens', in: Pratidanam. (Indian, Iranian
and Indo-European Studies presented to F.B.J. Kuiper on his sixtieth birthday, edited by J.C.
Heesterman et al. (The Hague/Paris: Mouton), 9-19.

Neu, Erich, 1974. Der Anitta-Text (= StBoT 18) (Wiesbaden: Harrassowitz).

Oettinger, Norbert, 1979. Die Stammbildung des hethitischen Verbums (= Erlanger Beiträge zur Sprach-
und Kunstwissenschaft 64) (Erlangen: Nürnberg).

Oldenberg, Hermann, 1967. Kleine Schriften (Wiesbaden: Steinert).

Pokorny, Julius, 1959-1969. Indogermanisches etymologisches Wörterbuch (Bern/Wien: Francke).

Pott, August, Friedrich, 1867. Wurzelwörterbuch der indogermanischen Sprachen (= Etymologische
Forschungen 3) (Detmold: Meyer).

Schrader, Otto/Nehring, Alfons, 1917-1929. Reallexikon der indogermanischen Altertumskunde
(2. Auflage) (Berlin/Leipzig: de Gruyter).

Schwyzer, Eduard, 1939--1950. Griechische Grammatik (München: Beck).

Szemerenyi, Oswald, 1964. Syncope in Greek and Indo-European and the Nature of the Indo-European
Accent (Napoli: Instituto Universitario Orientale di Napoli).

Thieme, Paul, 1949. Untersuchungen zur Wortkunde und Auslegung des Rigveda (Halle: Niemeyer).

100

Untermann, Jürgen, 1967. “Die Entdeckung des Genitiv singularis der -o-Stimme im Keltiberischen”,
in: Beiträge zur Indogermanistik und Keltologie, edited by W. Meid (Innsbruck: Sprachwissenschaftl.
Institut der Universitát Innsbruck), 281—288.

Wackernagel, Jakob, 1920-1924. Vorlesungen über Syntax 2 (2. Auflage) (Basel: Birkhäuser).

Wackernagel, Jakob, 1955. Kleine Schriften (Göttingen: Vandenhoeck & Ruprecht).

Wackernagei, Jakob/ Albert Debrunner, 1954. Altindische Grammatik 2/2 (Göttingen: Vandenhoeck &
Ruprecht).

Walde, Alois/Hofmann, Johann Baptist, 1938-1956. Lateinisches etymologisches Wörterbuch (Heidel-
berg: Winter).

Watkins, Calvert, 1969. Indogermanische Grammatik 3/1 (Heidelberg: Winter).

Genetic classification in linguistics and biology

Isidore Dyen (Yale University and the University of Hawaii)

The man we are honoring by this volume invited me to contribute to a seminar in which
biologists and linguists would join in comparing genetic classification in their respective
fields. To my great regret I was unable to accept. This article is my contribution to the
discussion of this subject.

Linguistic science and biology utilize the same family tree model. In biology the model is
used to diagram evolutionary developments leading from an elementary life-form to all of
the various members of the contemporary galaxy of life-forms. In linguistic science the
same model is used in diagraming the increase in membership in a family of languages. In
both fields the family tree diagram is thus a statement of a genetic classification based on
hypotheses of origins which are intended to explain data that are observable or otherwise
available in the form respectively of fossils and records. We will here try to explore some of
the similarities and differences that underlie the use of this model in these fields.

In biology it is commonly agreed to trace all life forms to a single origin. In linguistics it is
not common to trace all natural languages to a single common origin. The conclusion that
some languages are interrelated — that is, continue the same earlier language — is widely ac-
cepted if it can be reached by the linguistic comparative method which demands that
certain stringent criteria be met. It seems evident now that this method will not permit the
interrelationship of languages if their unity ended before a certain time, say before 15,000-
20,000 B.P.; many scholars might put this time much closer to the present. At the same
time it is generally accepted that even if the criteria of the comparative method are not
met, this fact does not exclude the existence of an interrelationship that the comparative
method cannot be used to reach, though, to be sure, such indeterminable interrelationships
are presumabiy always earlier than determinable ones. The comparative method require-
ments have only a positive effect, not a negative one, except for the range over which they
produce positive results.

Although failure to satisfy the requirements of the comparative method is not evidence of
unrelatedness, it is common parlance among linguists to speak of languages as being un-
related. This can occur when the comparative method has not given positive results. Thus it
is conceivable that languages commonly spoken of as unrelated — instead of, more properly,
unrelatable — might nevertheless be related (or interrelated).

In fact there is now good reason to believe that all human (natural) languages have a com-
mon source. All varieties of the human species are language-possessing and no other species
has what can be said to resemble a human language either in structure or complexity of

102

semiosis. The simplest hypothesis regarding the differentiation of the human species from
its most recently associated relative is that the differentiation was interdependent with the
development and exercise of a language capability. It seems hardly likely that this coin-
cidence of speciation with language acquisition could have occurred more than once in-
dependently. This proto-language of the species can be aptly called Proto-Human, a term
which is not my invention and already has some currency. (I should add that prior to my
realization of the need to give weight to the congruence of species with language-possession
I was skeptical of the utility of the Proto-Human hypothesis.)

The language type of Proto-Human is perhaps best regarded as being like that of any of the
languages that exist today, at least in respect to the factor of equicomplexity. There is a
growing tendency to recognize that in some sense all of the world’s languages are equal in
relation to the task of serving as a means of everyday communication. This recognition can
safely be extended to the proposition that any natural language can be brought to the
point of serving the same purposes as any other merely by expanding its vocabulary. A
possible formulation of this relation between natural languages is that all (human) lan-
guages are structurally equicomplex. This theory is justified by the observation that normal
individuals become minimally equipped language-manipulators in their respective societies
at about the same rate, that is, at about the same age. This observation implies that if we
assume that infants show the same average resistance to (or aptitude in) acquiring the struc-
ture of the language of their environment, the languages offer about the same degree of dif-
ficulty to the infants. This state of affairs can be attributed to the tension between the in-
creasing expertise of older speakers (parents, siblings, and others in the child’s environ-
ment) and the natural absorptive limitations inherent in the childhood state. Although the
power of acquiring a language structure distinguishes the human being, the actual exploita-
tion of this power demands time. It is interesting that an approximate equality in the rate
of acquisition is observable among children facing very differently structured languages.

It is hard to believe either (1) that the current universal level of language complexity was
reached prior to human speciation and was, as it were, a precondition for human specia-
tion, or (2) that human speciation was a prior condition for the initiation of the first step
toward language acquisition. It seems more in keeping with the gradualism that permeates
evolutionary evidence to imagine that the very first steps in language development above
that which can be observed among the other highest primates occurred in some variety of a
primate species which speciated as its means of oral intercommunication advanced, but
significantly long before the current level of complexity was reached.

After speciation the evolution toward our current level of language can be regarded has
being fostered by the advantages given by their uniquely possessed means of cooperation
and of storage and transmission of experience. It is conceivable that as the early language
form evolved, dialects developed with the increase of population. The hypothesis that such
dialects lie at the basis of our current unrelatable families seems prima facie to be tenable.
Under this hypothesis the dialects would have differentiated to the point of being no
longer recognizable as the same language; the requisite number of ancestors for our deter-
minable families could be reached either directly from the dialectalization of the original
primitive language or by the later branchings of the languages that resulted from the initial
subdivision. This hypothesis suffers from the weakness that it either (1) necessitates the
assumption that progression toward the current level of complexity is inevitable, since it re-
quires that some number of now different primitive languages reached the same level of

103

complexity and then stopped complicating, or (2) forces the abandonment of the equi-
complexity theory itself, if the level of equicomplexity is not inevitable.

If it be posited on the other hand that there was a single primitive language that progressed
to the level of complexity now evidenced by all current languages and that it was the suc-
cessive branching of this more complex language that ultimately led to our current lan-
guages, we have a simpler hypothesis that is equally satisfying.

This reasoning completes the argument for a Proto-Human which was equicomplex with
observable languages. At the same time this hypothesis leaves room for the notion that
what evolved into Proto-Human was not as complex as current languages. We shall refer to
the stages of Pre-Proto-Human after speciation that led to Proto-Human as Primitive
Human.

The argument for a Proto-Human presented here should not be regarded as of and by itself
justifying the Proto-Human that has been inferred by others from evidentiary arguments
based on collections of putative cognate sets. Such sets do not meet the comparative method
requirements and there is reason to believe that they can never meet these requirements.

The comparative method requirements for establishing relationship are very strict in the
hands of established comparatists. Some might wish to state that a proposed relationship
must be shown to be beyond doubt. Since a proposed relationship is a hypothesis by virtue
of being a hypothesis of common origin, it cannot be beyond doubt; a hypothesis beyond
doubt is a contradiction in terms. If the requirement is relaxed to read ‘beyond reasonable
doubt’, the issue becomes one of calculating probabilities, for the reasonableness of what is
reasonably doubled can be increased by the consideration of additional evidence. Unfortu-
nately no satisfactory way of calculating the probabilities involved in a primary relation-
ship hypothesis has been proposed, despite the fact that there is a consensus regarding a
large number of primary relationship hypotheses. The strictness of the comparative method
requirement seems thus to reside in the fact that any new relationship hypothesis must
meet the same requirements as some established (= accepted) hypothesis, for the latter
provides the only guidelines. It seems hardly likely that the current extensive and intensive
examination of possible but unestablished relationships will not soon eliminate all those
which satisfy this requirement, leaving some considerable number of unrelatable language
families.

The hypothesis of a Proto-Human does very little to assist in the establishment of the
family tree nodes leading to our observable languages. The most that it can do is to en-
hance the probabilities associated with putative cognate sets on the basis of which claims of
new relationships are based. If there was a unified Proto-Human, it is reasonable to believe
that the probability that putative cognate sets that look attractive include some really
cognate sets is greater than if such a language never existed. At the same time the degree of
enhancement is dependent on the estimate of the interval between Proto-Human and the
earliest proto-language established through the comparative method. The shorter the inter-
val, the greater the enhancement of the probabilities and, of course, vice versa.

If for the sake of argument the date of Proto-Human is set at 100,000 B.P., the interval
approximates 80,000 years or about four to five times the interval over which a relatively
optimistic view of the power of the comparative method to establish relationships claims
that it operates. It is hardly within the range of reasonable expectation to anticipate that
extensive and intensive work on currently established proto-languages might ever bridge

104

such a gap. The chances are obviously better if the dissolution of Proto-Human is set at a
more recent date; the collection of cognate sets would then appear to have better chances
of containing some truly cognate sets. However it also appears inevitably true that the
closer the date of Proto-Human is set to the present, the greater the difficulty faced by the
hypothesis is in explaining the extreme divergences among the world's language families.

It is intuitively satisfying to regard the child's acquisition of language as roughly recapitu-
lating the evolution of language through primitive stages to the present equilibrial state that
we have termed equicomplexity. There can hardly be any doubt that the phonemic prin-
ciple distinguishes human languages from animal intercommunication and characterized the
initial steps above the latter from the very start. This principle can be seen as permitting
the expansion — gradual or explosive — of grammatical structures to reach the plateau
represented by equicomplexity. Thus it would appear to follow that the child's acquisition
of the phonemic principle before the acquisition of grammatical intricacies resembles such
a recapitulation. Another example, trivial though it may be, is the acquisition of words
before word combinations. Other examples of recapitulations are bound to be discovered
as the typological (i.e. cross-linguistic) analysis of the child's acquisition of language ad-
vances.

The time between the acquisition of the phonemic principle by the Pre-Proto-Human and
the attainment of the highest complexity plateau (7 the level of equicomplexity) by the
Primitive Human must have been relatively short, for if it were long, the population could
be expected to have increased to the point of necessitating separations that might result in
at least markedly different dialects if not different languages. It is difficult to believe that
the entropy of the intercommunicational system characterizing all of the languages that we
know would not have begun to operate immediately. The rate at which the homogeneity of
idiolects in languages is dissipated depends on the rate of intercommunication between
contemporaries. As this rate falls toward zero, the rate of divergence increases, and at zero
the rate of divergence is maximum. Thus the longer the duration of Primitive Human, the
farther its dialectalization would have proceeded toward the outcome of different lan-
guages. Although this outcome can not be ruled out, the hypothesis of a unified Proto-
Human appears to require that the time for Primitive Human be both significantly long and
relatively short, say, at most a thousand to fifteen hundred years.

It seems both convenient and reasonable, though not at all satisfying, to regard the child as
having acquired a language at the moment that he has acquired a phonemic system ap-
propriate to that language. The child probably acquires the phonemic principle during his
second year and attains a complete phonemic system during his fourth year or shortly
afterward. Roughly speaking then, the 'gestation' period of a language takes about two
years, followed by years of increasing expertise and decreasing rate of advance. From this
point of view, the time needed for a child's acquisition of a language — viewed as a recapi-
tulation — is proportionately a larger percentage of the duration of Primitive Human than
the percentage that human gestation — viewed as a fetus's recapitulation of human evolu-
tion — is of the time over which human evolution took place.

Although without the phonemic principle grammatical complexities were probably un-
achievable originally, it does not follow that the child must acquire the complete phonemic
system of his language before he acquires any of the grammatical structure. Rather one can
expect the notion of arbitrarily different sounds as distinctions to appear before gram-

105

matical features are acquired. The child learns to deal with words, or what are effectively
words, before he learns ways of combining them. The distinction of grammatical features
can thus be regarded as in general dependent on the prior acquisition of a phonemic system
of some size, minimal though it may be. Nevertheless we can also expect some grammatical
features to appear before the child has completely attained his version of the phonemic
system of his language.

Within our present framework, as the theoretical moment of completing the acquisition of
a phonemic system approaches, the semasiological equipment of the child, on the average,
is in some sense about the same for all languages. This implies that their vocabularies are
about equivalent. It may very well be that the corresponding equipment of adults differs in
societies differing in development; this is indeed likely to be the case in view of the aids
and requirements that developed societies present to their members. However since children,
if unhampered by the debilitating effects of disease, malnutrition, and neglect (theoretical-
ly irrelevant and thus discountable despite their practical effects on an average) can be
expected to pursue the acquisition of language competence at about the same rate, their
semasiological complement should also advance at about the same rate. The apparent ad-
vantage offered by languages with smaller phonemic inventories, which might appear to
permit the semasiological component to increase more rapidly, is offset if — as we have
some reason to believe — smaller phonemic inventories require longer meaningful units.

The phonemic principle lies at the very heart of the comparative method. The comparative
method is based on the law of regular phonetic change which requires that in continuing
words a phoneme can change in different ways only in differing phonemic environments.
Associated with this law is the law of dialect cohesion which requires that all pairs of
idiolects (or dialects) of the same language should show large numbers of approximately
homosemantic words whose phonemes correspond systematically and are most heavily
distributed among the words of highest frequency in both members of the pair. The con-
sequences of the law are that dialects of the same language universally exhibit large numbers
of words with the same or similar meaning whose phonemes correspond systematically, a
relation which can be regarded as facilitating the intercommunication between speakers of
phonemicaliy differing dialects.

Extra-legal changes involving phonemes are in general sporadic and can nearly all be clas-
sified as analogical changes or spurious because they involve borrowing. Since borrowing
occurs not only between different languages, but also between different dialects of the
same language, a given dialect may show sets of doublets that differ regularly with each
other.

The equicomplexity of language structures does not appear to have an analog in outcomes
of biological evolution. One might look for it in local adaptation to the environment, for
each species almost universally seems admirably adapted for survival in its environment and
must be so on pain of extinction. Nevertheless unicellular animals are inescapably less
complex than multicellular animals. Furthermore the differences in complexity of brain
structure found in higher animals can hardly be equated with differences in structural
complexities elsewhere in lower organisms.

The dialectological language appears to be a nice analog of a biological bisexual species. A
dialectological language consists of all idiolects (or, it one wishes, dialects) connected by a
chain of mutually intelligible pairs of native speakers. Similarly a bisexual species can be

106

defined as consisting of all varieties (or individuals) connected by a chain of pairs of viably
reproductive individuals (in the sense of being able to produce fertile offspring). A chain of
pairs as used here is a set of all pairs (1) arranged so that each individual is a member of as
many pairs as there are other individuals with whom it shares the defining criterion and (2)
connected together by an imaginary line (or curve) which can pass through all individuals
regardless of geographical distribution by passing from one pair to another only through a
common member, retracing being permitted.

Such an arrangement seems to suggest the possibility that, within the set of individuals who
can interact, a minimum rate of change is operative so that though local homogeneity is
maintained, diversity develops among non-interacting individuals. When the individuals of
different groups interact in reduced degree or not at all, the changes affecting their respec-
tive dialects tend to be independent. It is the rate of interaction — in the case of languages,
intercommunication — that determines both homogeneity and diversity within a language,
but only because of the constancy of change. The fact that languages behave according to
this model is exemplified by all languages whose history can be traced through records. All
have changed in some degree even when the elapsed time is only a few hundred years. At
the same time, obviously, no difficulty in intercommunication arose locally, whereas
among groups who for some reason were separated, great differences have arisen, some-
times, as in the case of Rumanian which became separated from the rest of the dialects
that continued Latin, leading to the rise of different languages. In this theory of a mini-
mum rate of change there is at least a partial explanation of the fact that all languages ex-
hibit dialectalization.

This model does not fit evolutionary developments in biological evolution so well if it is
true that some species have been able to persist essentially unchanged over extremely long
periods of time. A fossilized ant from a period in the far distant past is reported to be in-
distinguishable from a contemporary species. At the same time it could be true that there is
a difference in this respect according to position on the evolutionary scale, so that the
higher species are subject to a minimum rate of change, whereas lower species are not.
Perhaps it is also of some importance that asexual species have no analog among languages.

The manner of language increase and that of species appears nevertheless to be essentially
the same. Both result from the increasing divergence between unispecial non-interacting
varieties. A final split can not occur between interacting varieties. The final split can thus
occur only (1) if parts of a chain become isolated from each other so that no interaction
takes place between any members of the separated parts for the length of time necessary
for changes on both sides to destroy the possibility of interaction or (2) if a linking group
between extremely divergent parts of a highly diversified chain is wiped out so as to leave
only groups that can not interact with each other.

Such an increase has a superficial resemblance to mitosis since what was a unified whole
becomes separate units. Although linguists appear to permit splitting simultaneously into
more than two branches at once, they do so when — as happens in many instances — their
evidence does not allow distinguishing the sequence of some number of divisions. This
might result from splits which occur too close together in time to give sufficient evidence
to determine their sequence securely. Such supraunitary splits are permitted in the bio-
logical realm for the same reason.

107

In linguistics the dominant trend among comparatists is to insist on the non-merging prin-
ciple. They assume that languages do not mix, at least in the sense that a language bound-
ary, viewed as what occurs between dialectological chains regardless of geographical distri-
bution, can not disappear except with the demise of one of the bounded languages. This
principle can only be maintained under a definition of ‘a language’ that is essentially like
the one presented above for a dialectological language and if it is also recognized that a
creole language like Takitaki (qua dialectological language, not to be confused with a creole
dialect of some dialectological language like Jamaican English) is not a-native-speaker-to-
native-speaker continuation of any other languages or language, for that matter, and is thus
a newly originated language. The native idiolects associated with it from the time it ac-
quired native speakers were never part of any other chain. In this way, though a creole lan-
guage may satisfy the notion of mixture under some definition of mixture, it does not
exemplify language merger in any sense that affects the traditional notion that a language
boundary that has once appeared cannot disappear through such an event.

The non-merging principle fits with our intuition as to what can be expected to happen if
people speaking different languages begin to interact. Normally an agreement can be ex-
pected to be reached as to which language will be used with each other by members of
different groups. Though loanwords from the other language can be expected to be brought
into the agreed-upon language, the integrity of the latter would not be affected. Creole
languages are best regarded as a product of the special conditions that arose during the
post-Columbian colonial period since there is no indication that creole languages were
produced in slave-collecting empires before that time.

The importance of the non-merging principle for making inferences about the past of lan-
guages is that it serves to simplify the task of establishing and reconstructing proto-lan-
guages. As a result of the operation of the law of regular phonetic change, cognates (i.e.
words that continue the same word from the common proto-language) show the same sort
of systematic correspondences that are shown by so many words of the same or similar
meaning in different dialects of the same language. To be sure the phonemic correspon-
dences between words of different languages involve differences of greater complexity than
those shown by words of different dialects of the same language. Nevertheless it is just such
systematic correspondences that permit not only the identification of related languages,
but also directly the reconstruction of the phonology of the proto-language continued by
the related languages — each in its own way — and indirectly many of the other features
of the proto-language, such as at least in part, where not fully, its structure in the form of
its morphology and syntax as well as even the meaning of some of the words that can be
reconstructed. If languages could merge, such inferences would be difficult if not im-
possible.

It is perhaps worth observing at this point that the non-merging principle applies to what is
expected to occur in nature. One can imagine that with a great deal of effort a connection
over gradually varying dialects could be induced artificially, just as, I believe, it is possible
to manipulate the genes to produce interspecial varieties which would then constitute a
connection between what had been before regarded as different species. Our definitions
apparently cannot be expected to operate in the face of artificial interference.

I am not familiar enough with the course of reconstruction from the comparison of oc-
current forms in biology to do better than to comment that it is my general impression

108

that the reconstruction of a proto-creature through this type of inference is not generally
engaged in. There is no clear analog of the law of regular phonetic change, but there is
almost obviously strong reason to hope that the necessary information will prove attain-
able from the triads of the helixes to ultimately provide the steps by which the character-
istics of ‘missing links’ can be inferred and even the chronological periods of their existence.

Some notes on Dravidian intensives*

Murray B. Emeneau (University of California, Berkeley)

Reduplication of forms in the Dravidian languages has not yet received full treatment,
either descriptive or comparative-historical. Nor does it seem probable that the latter will
be possible in any productive sense until full descriptive treatments of at least a majority of
the languages are available.

It is already evident, however, that several different semantic categories are involved in this
type of morphology, two important ones being distribution (‘street after street’, ‘ten each’,
‘each by itself”) and intensity (‘high degree’, ‘excellence’, ‘emphasis’, ‘very’). A third cate-
gory which involves reduplication consists of expressives (onomatopoetics). In this latter
category it has already been shown that the morphology sometimes has instead of identical
reduplication some substitution for part of the basic form (e.g. Ta. katapatav-enal *hullaba-
loo") (Emeneau 1969: 274—299). A semantic category which is always expressed by redu-
plication with substitution is that of the ‘echo-words’ (‘A and the like”, ‘A etc.’; the forma-
tion: CVX giX) (Emeneau 1938: 109--117).

Pending the desirable full treatment, which is probably realizable only in the distant future,
this paper calls attention to some details of the intensive formation.’ No full treatment is
yet at hand for any of the languages. The dictionaries are hardly exhaustive, and the ety-
mological dictionary (DED, etc.) did not succeed in recording everything that was avail-
able. Kittel’s description of the formation in Kannada is perhaps as good as any, with
numerous examples (from the literature) divided by ‘parts of speech’; e.g. huli huli ‘very
sour (of buttermilk)’, heccu heccige ‘more and more, excessively’, bela belagu ‘excessive
lustre’, piri piridu ‘that is very large’, Kappa kappane ‘very black’. The Tamil Lexicon con-
tains many examples; e.g. veluvelu ‘to become white, pale’, velu-velu-enal ‘extreme whit-
eness or paleness’, veluveluppu ‘whiteness, pallor’ (: vel ‘white’, velu ‘to become white,
pale’); kKarukaru ‘to become very black’ (: karu ‘to grow black’); futituti ‘to be in a great
flurry, fret and fume’ (: tuti ‘to be in great flurry or agitation, tremble, palpitate’); cilircilir
‘to get goose-skin from intense emotion’ (: cilir ‘to bristle as the hair on the body’);
kuraikurai ‘to be confused’ (: kurai ‘to be troubled’); karakara ‘to feel irritation (in eye,
throat)’ (: kari id.). Telugu has the formation; e.g. ceraceral-àdu ‘to be angry or furious’;
korakora ‘anger; angrily’, korakoral-adu ‘to be angry, have an angry look’. Examples are
found in the descriptions of the non-literary languages of the center and the north, but there

* [n this volume honoring Henry Hoenigswald on his 70th birthday, 1 add my homage to one who has
notably advanced comparative-historical linguistic methodology. My paper illustrates how abortive
such a study may be when we lack adequate descriptive data, but also, I hope, that to try to point
out a direction for further research may promise future profit.

110

is no satisfactory description, and such an example as Kur. bilbilrna ‘to sparkle, twinkle’
(: bilcna ‘to shine, glitter, sparkle’), Malt. bilbilre ‘to shine brilliantly’, alongside Ta. veluvelu,
etc. (quoted above), Ma. veluvela ‘very white’ (: velu- ‘to be white’), can hardly be evalu-
ated.

Kittel defined the intensive reduplication in Kannada as meaning: “to express high degree,
excellence, intensity or emphasis” (1903: 302ff., §§ 303.1(d), 2(a), 305(a), 307(a)), also
using ‘very’ commonly in his translations of examples.” Arden (1934: 283, § 579(v)) on
Tamil says: “the first part is reduplicated to intensify the meaning, as natunatunku tremble
greatly, a reduplicated form of natunku tremble’. Such forms in the Tamil Lexicon are
sometimes translated with an indication of this meaning, sometimes not. Of the several
forms connected with velu ‘to become white, pale’ that were quoted above, veluvelu is
given with the same meaning as velu, but velu-veluv-enal is defined as ‘extreme whiteness or
paleness’. Of the other forms quoted above, karukaru ‘to become very black’ is so defined
as contrasted with karu ‘to grow black’; but for none of the other reduplicated forms is
there given a meaning that explicitly includes 'intensity'. Similarly with most of the other
reduplicated forms that could be quoted; the simplex form and the reduplicated form are
usually given identical or almost identical meanings, and we must conclude, by extra-
polation from the few examples in which the explicit Kannada meaning of 'intensity' is
paralleled, that the compilers of the Tamil Lexicon have been somewhat inexact in the
meanings they give.

Once the intensive formation is established, it is possible to use it as the basis for some
formations whose meaning is ‘maximal intensity”. The difficulty caused by incomplete
recording of forms and defective dictionary meanings is particularly evident here. Very few
contrasting pairs of forms are available in the material at hand. Entry 1 below, Ta.
puttapputiya, puttamputiya ‘brand-new, very recent’ (: putu, putiya ‘new’), establishes
the meaning; similar are entry 4 Ta. verra-veritu ‘absolute worthlessness’, entry 12 Ta.
tannan-tani ‘quite alone’, entry 7 Ka. kattakade, Te. kattakada ‘the very end’; for none
of these is there recorded an intensive formation (*veraveritu, etc.). A few examples
have contrasting pairs, but in no case is there recorded contrasting meaning; e.g. entry
14 Ta. veluvelu ‘to become white, pale’, velu-veluv-enal ‘extreme whiteness or paleness’:
vella-vilar ‘to become very white (as a washed cloth)’ (: vilar ‘to become pale, whiten’ =
velu); see also entries 3, 5, 15, 16,17.

Clear examples yield the following statement of formation: in C!vC2V of the initial
member of the reduplicated form, c2 being a coronal (Schiffman 1975: 69—85;
1980: 101—110) consonant (but not c),

voiced stops d rd -* voiceless stops tf rr ft (in Tamil-Malayalam the voiced stops are
written t r f);

nasals n n > nn na;

lateral / > IL

The largest number of examples has been recorded for Tamil. There are examples also for
Kannada, Telugu, and Kota. Some etymological groupings are found in the material, but
the almost complete certainty that the material at hand is only a sample makes it impossible
to attempt comparative-historical statements which would settle such questions as whether
we can reconstruct the formation for South Dravidian (or general Dravidian?), or whether
there has been extensive borrowing from one language of origin (e.g. Tamil).

111

The entries for the ‘maximal intensive’:

*d > *tt

(1) Ta. puttapputiya, puttamputiya ‘brand-new, very recent’ (: putu, putiya ‘new’) (3511/
4275)

(2) Ka. tuttatudi ‘the very point or end’ (: tudi ‘end’; cf. Ta. tuti ‘point’, Te. tuda, tudi
‘end’) (2716/3314)

(3) Ka. mottamodalu ‘the very beginning, at the very first’ (: moda-modalu, modalu
modalu id., modal(u) ‘state of being first?) (4053/4950)

“r > “rr

(4) Ta. verra-veritu ‘absolute worthlessness' (: veritu *uselessness, futility', veru ‘empty,
useless, etc."), Ko. veta’ ver ‘mere, unmixed, vain' (: ver id.) (4538/5513)

*d > *tt

(5) Ta. natta-natu ‘the very middle’ (: natu ‘middle’), Ka. natta-naduve ‘in the very middle’
(: nadu-naduve id., nadu ‘middle’, adj. natta), Te. natta-nadumu ‘the very middle’ (: nadu
‘middle’, adj. natt-), Ko. nata’ narv ‘very center’ (nary ‘middle, center’) (2959/3584)

(6) Ta. netta-netumai ‘great length, excessive tallness or height’ (: netu ‘long; to be long,
tall’, netumai ‘length, tallness, height’, nettam, nettu id.) (3099/3738)

(7) Ka. kattakade ‘the very end; at the very end, at last’ (: kade ‘end, last’), Te. kattakada
‘the very end, the very last place or point; farthest, hindmost, the very last’ (: kada ‘end,
extremity’) (929/1109)

(8) Ko. cet ceyr ‘bad smell’ (: ceyr, cet id.; cf. Ta. ceti ‘stench’, Tulu sedi ‘strong smelling’,
Settuni ‘to decay’) (1614/2760)

(9) Ko. muta’ murmurn ‘without any reason’ (: murmurn id.; cf. Kodagu mundate ‘for no
reason’) (4037/4928)

(10)? Ta. vettaviti ‘break of day, early dawn’ (: viti ‘dawn; to dawn’, vettam ‘light’, vetta
‘clear, plain’; cf. vetta-veli ‘open plain’, vetta-veliccam ‘broad daylight’). Perhaps a com-
pound with vetia. (4504/5475)

*n > "nn

(11) Ta. inn-ini ‘now, without a moment’s delay’ (: ini ‘now’) (351/410)

(12) Ta. tannan-tani ‘quite alone’ (: tani ‘solitude; to be alone’) (2612/3196)

*n ^ *nn

(13) Ta. annani ‘in close proximity’ (: ani ‘near’) (69/120)

* > Il

(14) Ta. vella-vilar ‘to become very white (as a washed cloth). (: veluvelu ‘to become white,
pale’, velu-veluv-enal ‘extreme whiteness or paleness’, vilar ‘to become pale, whiten’, vel
‘white’, velu ‘to become white or pale’; cf. Ma. veluvela ‘very white’, Kur. belbelrna ‘to
have an unusual or morbid whiteness of the skin, be an albino’, bilcna ‘to shine, glitter,
sparkle’, bilbilrna ‘to sparkle, twinkle’, Malt. bilbilre ‘to shine brilliantly’) (4524/5496)
(15) Ta. kulla-kkulir ‘to be intensely cool and refreshing’ (: kulir ‘to be cool, refreshing’; cf.

Ma. kulukulu ‘intense cold’, Ko. kulkul in- ‘to feel cool, feel calm and peaceful’) (1523/
1834)

112

Two other sets of forms may belong here, with rules

*r> *nn and #1 > *nn

(16) Ta. pennam-periya, pennam-perutta ‘very large’ (: periva, perutta ‘large’, [Caldwell
quoted above] periya periya ‘very great’; cf. Te. penu [before consonant], penn- [before
vowel ‘large’) (3613/4411)

(17) Ta. punnam-pulari ‘early dawn’ (: pulari ‘dawn’, pular ‘to dawn’, pur-pull-enal ‘break
of day’; cf. Malt. pulpulre ‘to shine through whiteness’) (3531/4305)

No instances have been found involving non-coronal consonants (labial, velar), nor the
palatal c. The following unique set of forms is very puzzling, but may possibly belong here.

(18) Ta. cekka-cciva, cekkañ-ceku ‘to be deep red’, cekka-cciver-enal, cekka-ccevér-enal
‘deep redness’ (: civa ‘to redden’, civér-enal, cever-enal, cekkam ‘redness’, but no *ceku/)
(1607/1931)

Two sets in Kannada and Telugu have items in which
*d > *tt, instead of *d > *tt. In one set Kannada has both forms, the expected *tt having
been given above as entry 2. Nothing further of this kind has been found.

(19) Ka. tuttatudi ‘the very end or point’ (cf. tuttatudi in entry 2, Te. tuttatuda ‘the very
end or extremity’ (2716/3314)

(20) Te. motta-modalu ‘the very beginning’, motta-modata(n) ‘at the very beginning’
(: modalu ‘beginning’, modata(n} ‘at first, in the beginning’; cf. Ka. motta-modalu in entry
3 (4053/4950)

Several formations which show some similarities to those already treated are to be analyzed
as compounds. Entries 21 to 23 have nothing to do with 19 and 20 in spite of ff in the first
elements.

(21) Ta. kattu-kkaval ‘strict guard’, Ko. kat-ka'vl ‘close watch’. The first element is Ta.
kattu ‘firmness, strength’ (962/1148); kaval ‘guard, watching’ (1192/1416). Kota has un-
doubtedly borrowed from Tamil.

(22) Ka. kotta-kone ‘extreme point’, Te. kotta-kona ‘the very end or extremity’. *kottu
‘tip, point, crest’ (1704/2049); *konay ‘tip, point, extremity’ (1807/2174).

(23) Ta. patta-ppakal, Ma. patta-pakal, Te. patta-pagalu ‘broad daylight, midday’, To. of
oxol ‘broad daylight’ (used in song). *pakal ‘day(time)’ (3151/3805); with *patta-, cf. Ko.
pata'r in- (country) becomes fully light at dawn’. The Toda form has the earmarks of a
borrowing from Kannada (Badaga dialect), i.e. *p- > Ka. h- > To. ¢, but the source form
has not been recorded so far.

(24) Ta. cinnafi-ciriya ‘very small’. The Tamil Lexicon calls this reduplication, but it is
rather a compound of cinna (2135/2594) and ciriya (1326/1594), both ‘small’; the com-
pound is not in the etymological dictionary.

(25) Te. nirra-nilugu ‘to be very haughty’. Te. nilugu ‘to be conceited’ (3059/3692); for
nirra-, cf. niru ‘very, much’ (3049/3682).

113
Notes

Most of the data in the paper come from Burrow-Emeneau (1961), A Dravidian Etymological
Dictionary (DED), its Supplement (DEDS), and the 2nd edition revised (DEDR). A few items have
been added from other sources, especially Kannada items from Kittel’s grammar and dictionary.
The paper would have been bettered by consultation with speakers of the languages, but this could
not have been easily accomplished. — Each of the numbered entries has at the end a pair of numbers,
the first a reference to DED and DEDS, the second to DEDR. — Abbreviations of language names:
Ta. = Tamil, Ma. = Malayalam, Ko. = Kota, To. = Toda, Ka. = Kannada, Te. = Telugu, Kur. = Kurux,
Malt. = Malto.

Caldwell earlier had mentioned the construction in his treatment of the superlative of adjectives, as
‘the very primitive plan of doubling of the adjective itself — e.g. [Ta.} periya-periya, very great’
(1913: 317).

References

Arden, Albert Henry, 1934. À progressive grammar of Common Tamil (4th ed.) (Madras: Christian
Literature Society).

Burrow, Thomas/Emeneau, Murray B., 1961, 1968. À Dravidian etymological dictionary (+ supple-
ment) (Oxford: Clarendon Press).

Burrow, Thomas/Emeneau, Murray B., 1984. A Dravidian etymological dictionary (2nd ed.).

Caldwell, Robert, 1913. A comparative grammar of the Dravidian or South-Indian Family of Languages
(3rd edition) (London: Harrison).

Emeneau, Murray B., 1938. “Echo-words in Toda”, New Indian Antiquary 1: 109-117 (= Dravidian lin-
guistics, ethnology and folktales, 1967 (Annamalainagar), 37-45).

Emeneau, Murray B., 1969. “Onomatopoetics in the Indian linguistic area", Language 45: 274—299
(= Language and linguistic area, 1980 (Stanford), 250-293).

Kittel, Ferdinand, 1903. A grammar of the Kannada language (Mangalore: Basel mission book and tract
depository).

Schiffman, Harold, 1975. “On the ternary contrast in Dravidian coronal stops”, in: Dravidian phonolog-
ical systems (Seattle) 69—85.

Schiffman, Harold, 1980. “The Tamil liquids”, in: Proceedings of the Sixth Annual Meeting of the
Berkeley Linguistics Society (Berkeley: Berkeley Linguistics Society, University of California), 101—
110.

Vedisch dyavasa-

Bernhard Forssman (Universität Erlangen-Nürnberg)

$ 1. Ein Nominalstamm dyavasa- ist zuerst im Rgveda überliefert. Er kommt hier nur an ei-
ner Stelle vor, I 122,15. Der schwierige erste Halbvers lautet:

catvaro mà masarsárasya sisvas
tráyo rajfia ayavasasya jisnóh.

In Graßmanns RV-Übersetzung findet sich folgende vertretbare Wiedergabe: ‘Mich [fahren]
vier vom Sohne des Magargara geschenkte Rosse und drei des siegreichen Königs Ajavasa’.
Geldners Ubertragung weicht teilweise ab; sie ist deswegen einstweilen wohl nicht vorzuzie-
hen, weil er in sisvah nicht wie GraBmann eine Genitiv-Singular-, sondern eine ungewöhn-
liche — wenn auch nicht undenkbare — Nominativ-Plural-Form sieht. Uber avavasasya je-
doch hat Geldner dieselbe Meinung wie Graßmann: Auch er nimmt hier einen Personenna-
men im Genitiv Singular an! . Diese Bestimmung dürfte richtig sein; sie wird auch von ande-
ren Forschern vertreten und findet sich bereits bei Säyana.

Falls der Name Ayavasa- mit dem späteren homonymen Appellativum eins ist, so hat er ur-
sprünglich etwa ‘keine Weide habend, weidelos’ bedeutet ($ 2). Möglich ist ein Personen-
name mit dieser Bedeutung zweifellos. Als Stütze für diese Annahme kann — neben der
Form dyavasa- selbst — das vrddhierte Patronymikon Sauyavasi- (AB) angeführt werden. Es
dürfte nämlich zeigen, daß auch das Grundwort süyavasa- (88 4—5), das Gegenstück von
dyavasa-, als Personenname dienen konnte (‘mit guter Weide”). Doch muß eingeräumt wer-
den, daß Bildung und Bedeutung von RV dyavasa- nicht ganz sicher zu bestimmen sind.
Trotzdem ist der vereinzelte Beleg eine wichtige Stütze für die ohnehin wahrscheinliche
Annahme, daß das erst etwas später bezeugte appellativische dyavasa- in ältere Zeit zurück-
geht.

82. Dieses appellativische Zyavasa- ist unter anderem in der Taittiriya-Samhitä (TS) über-
liefert, V 2,8,3. Hier wird begründet, warum beim Bau des Feueraltars auch eine durvestaka-,
ein ‘Ziegel mit durva-Gras, Düurvä-Ziegel”? , hinzulegen ist:
pasüh | va esd vad agnir, nd khdlu vdi paséva äyavase ramante, dürvestakàm üpa
dadhati, pasunam dhrtyai

*Ein Tier ist ja Agni; Tiere verweilen ja bekanntlich dyavase nicht; man legt einen Dürvà-
Ziegel hin; das geschieht zur Bewahrung der Tiere’.

Der mit einem Tier identifizierte Agni wird sich also, so erwartet der Opfernde, bei dem
Dürvä-Gras behaglich fühlen?. Begründet wird diese Erwartung mit der in dem Satz nd

116

khálu vdi pasáva dyavase ramante ausgedrückten Lebenserfahrung. Da er die Satznegation
nd enthält, muß substantivisches dyavasa- hier eine Gegebenheit bezeichnen, die zwar etwas
mit ydvasa- Ntr. ‘Weide’ zu tun hat, aber bei Tieren kein behagliches Gefühl bewirkt*. Der
Kommentar übersetzt dyavase ansprechend mit alpaghase pradese “an einem Ort mit wenig
Nahrung’. Nur geringfügig anders lauten die Wiedergaben von Böhtlingk (Sanskrit-Wörter-
buch) und von Keith (TS-Ubersetzung): ‘Futtermangel’ bzw. ‘want of grass’. Diese Auf-
fassung ist ebenfalls möglich; da aber bei ram häufig Ortsangaben erscheinen, ist ‘weideloser
Ort, futtergrasloser Ort’ vielleicht vorzuziehen.

$ 3. Die TS-Stelle hat in den anderen Samhitäs des Schwarzen Yajurveda Parallelen. Dabei
treten gewisse Abweichungen in Erscheinung; der Satz mit ayavase jedoch lautet, von
Partikeln abgesehen, überall gleich. MS IH 2,6: 25,1 pasdvo vd istakà, nd vai pasdvà dyavase
ramante, ydd dürvestakäm upadddhäty äydtanam iva va etät kriyate, pasunam yáty(ai)
‘Tiere sind ja die Ziegel, Tiere verweilen an einem Ort ohne Weide nicht; dadurch, daß er
einen Dürvä-Ziegel hinlegt, wird ja gleichsam ein Standort gemacht; das geschieht zum Fest-
halten der Tiere". — KapKS XXXI 8 : 180,23 = KS XX6 : 24, 19 athaisa durvestaka/pasur
va agnih/na pasava dyavase ramante/pasubhya evaitad avatanam karoti/pasunam dhrtyai
‘nun der Durva-Ziegel; ein Tier ist ja Agni; nicht verweilen Tiere an einem Ort ohne Weide;
so macht er den Tieren einen Standort; das geschieht zur Bewahrung der Tiere’. — Aus ei-
ner weiteren Parallelstelle kónnte schlieflich das Zitat bei Patafijali V 4, 36, 6 (II p. 435,
22 Kielhorn) stammen: ayavase vardhante ‘(sie) gedeihen an einem Ort ohne Weide’. Viel-
leicht hat der Satz an der verloren gegangenen Stelle vollständig etwa so gelautet: na vai
pasava dyavase vardhante*’ . Falls diese Vermutung zutreffen sollte, könnten alle bisher be-
kannten Belege von ved. dyavasa- ein und demselben Sprichwort? zugerechnet werden.

$ 4. Die negative Bedeutungskomponente von dyavasa- deutet darauf hin, dafs in dem d-
irgendwie das Privativpräfix enthalten ist. Wenn dessen übliche Form mit kurzem a- ange-
nommen werden soll, so müßte das á wohl entweder durch Vrddhi oder durch die Lokal-
partikel @ zustandegekommen sein. Folgende zwei Wege kimen dann etwa in Frage: 1)
vavasa- > a-yavasa- > dyavasa- (mit Vrddhi); 2) yavasa- > a-yavasa- > (&- ‘in, bei’ + ayavasa-
2) ayavasa-.

Vrddhi von Privativbildungen (gemäß 1) kommt schon im RV vor (amitra- : amitra-); die
Annahme einer solchen Bildung kónnte insbesondere durch das parallele, nur wenig jüngere
und tatsichlich vrddhierte sauyavasa- abgestützt werden, das sogar in einem ganz áhnlichen
Wortlaut vorkommt, KB XI 5 am Ende (bisher die einzige Stelle): upanivartam iva vai
pasavah sauyavase ramante ‘irgendwie immer wiederkehrend verweilen ja Tiere in einer Ge-
gend mit guten Weiden’. — Auch ein Kompositum mit 4 ‘in, bei’ als Vorderglied (gemäß 2)
wäre immerhin denkbar, wenn diese Bildeweise in der älteren Sprache auch nicht sehr häu-
fig zu sein scheint”.

Beide Deutungen haben indes einen gemeinsamen Mangel. Sie setzen als Brücke zwischen
ydvasa- und dyavasa- eine einfache Privativbildung a-yavasa- voraus. Es ist nicht recht einzu-
sehen, warum sie trotz ihrer Einfachheit nicht literarisch überliefert ist!°. Dabei kann eine
solche Privativbildung angesichts der guten Bezeugung des einfachen sü-yavasa- (RV +; der
Akzent schwankt) durchaus erwartet werden. Ist doch su-yavasa- mit seiner Bedeutung
(‘mit guter Weide’, ‘gute Weide”) etwas wie ein Gegenstück zu dyavasa- und enthält sicher
nur die zwei Elemente su ‘gut’ und yavasa- ‘Weide’.

117

§ 5. Gerade suüyavasa- mit seinem Wechsel zwischen sū- und su- eröffnet aber einen ande-
ren Weg zur Erklärung von ayavasa-. Das auffällige sü- ist hier eindeutig die ältere Form des
Vordergliedes; es gilt im RV für sayávasa- und seine Weiterbildungen!! durchaus. Der Pa-
dapätha hat überall su- durchgeführt. Auch die einzige Belegstelle in der TS (1 7, 5, 2-3 bis)
zeigt im Text das rgvedische suydvasa-, ersetzt durch su?^ im Padapátha. Original belegt ist
das jüngere suyavasa- erst ŚB XI 7,1,1 (bis)? und BSS XXVII 14 : 341,1. Es kann leicht als
vereinheitlichte Form erklärt werden; als solche ging es auch in den Padapätha ein'?.

$ 6. Weiterhin zeigt sich Länge des Fugenvokals in ku-yava- TS IV 7,4,2, einem Komposi-
tum aus pejorativem ku- (ursprünglich ‘wo?’) und ydva- ‘Gerste’, dem Grundwort von
yävasa-. Doch ist küyava- aus zwei Gründen für eine Beweisführung hier weniger gut ver-
wendbar als suydvasa-. Einerseits ist die Form mit Kurzvokal, also kuyava- “Mifernte’, frii-
her überliefert (RV); küyava- erscheint außerhalb der TS!* nur als Handschriftenvariante in
der Parallelstelle MS II 11,4 : 142,1. Andrerseits gibt es neben ku(-) auch sonst Formen mit
ù, z.B. kú cit ‘überall’ (RV).

$ 7. Die Fugenvokaldehnung in suydvasa- ist, gemäß einer verbreiteten Anschauung, laryn-
galbedingt'^. Gelten dieselben Bedingungen auch für dyavasa-, so enthilt dieses als Elemen-
te nur das Privativpráfix und yávasa- und ist damit eben das gesuchte einfache Gegenstück
(8 4) zu suyavasa-. Das ist sicher eine morphologisch und lexikalisch befriedigende Erklä-
rung.

$8. Zu besprechen bleibt die Länge des @-. Ein brauchbares und schon bekanntes Parallel-
beispiel für anlautendes g- < n-a- (wie in jata- ‘geboren’ < *Ena,-to-) ist asant- ‘nicht seiend,
nicht wahr’, das im RV fünfmal und nur in alten Büchern deutlich überliefert ist (IV! V!
VII?). Regelmäßig aussehendes ásant- liegt im RV ebenfalls an fünf Stellen deutlich vor,
von denen aber drei in einem jungen Buch stehen (VI! VII! X?); zwei weitere Belege kón-
nen wegen des Sandhi dsant- oder dsant- enthalten (VII 104, 12b; X 129,1).

Graßmann weist in seinem Wörterbuch darauf hin, daß 2- zweimal an Versstellen steht, die
eine Kürze begünstigen. Der Padapätha hat an sämtlichen Stellen d- durchgeführt. In späte-
ren Texten kommen offenbar keine selbständigen Belege für dsant- mehr vor. Bei dsant- hat
die weiter abliegende Bedeutung ‘unwahr’, die bei sicherem dsant- im RV fehlt, das Überge-
wicht!$. Somit spricht tatsächlich eine ganze Reihe gewichtiger Gründe dafür, daß die Pri-
vativbildung äsant- etwas Altertümliches ist. Die Herleitung von dsant- aus *n-2, sont- ist
Kuryłowicz zu verdanken”.

§9. Die allmähliche Ersetzung von äsant- durch äsant-"? könnte darauf deuten, daß auch
in anderen Fällen der Anlaut a- von Privativbildungen einem a- Platz gemacht hat. Anwärter
für ehemaliges 4- sind u.a. vielleicht die im RV bezeugten Stämme a-dánt- ‘zahnlos’, a-renú-
*staublos', a-vatá- *windlos'. Doch ist stets auch die Móglichkeit in Betracht zu ziehen, daß
die Bildung erst erfolgte, als das zweite Element (also dant- ‘Zahn’ usw.) den anlautenden
Laryngal bereits verloren hatte.

$ 10. Erwogen werden muß auch, ob das privative d- von dyavasa- nicht durch sekundäre
Ersetzung zustande gekommen ist, z.B. durch Übertragung aus lautgesetzlichen Fällen wie
dsant-?, Eine solche Ersetzung scheint mindestens in einem Fall tatsächlich stattgefunden

118

zu haben, in ddeva- ‘gottlos’, das RV II 22,4, VI 49,15, VIII 59,2 (Khila) überliefert ist und
die gleiche Bedeutung wie häufiges ddeva- hat?°. Wie das é von ddeva- zu erklären ist, ist
nicht ohne weiteres ersichtlich. Ausgangspunkt Könnte der Vers VI 49,15 sein, in dem kurz
nacheinander ddevih und ädevih vorkommen. Oldenberg (Noten, zur Stelle) spricht von ei-
nem “Streben nach Abwechslung”; derlei mag öfters mitwirken, doch wird es kaum je die
einzige Ursache sein?! .

Nun macht das ddeva- der drei RV-Stellen einen eher gekünstelten Eindruck. dyavasa- ge-
hórt dagegen der Prosa an und wirkt nach Bedeutung und Kontext ($8 2—3) wie eine Vo-
kabel der Normalsprache. Eine Einwirkung von dsant- oder von ädeva- läßt sich kaum wahr-
scheinlich machen; und weitere klare Fälle mit deutlich sekundärem privativem d. wie
ddeva- scheint es im Vedischen nicht zu geben. Was sonst dafür beansprucht wurde??, ist
entweder ganz unsicher wie érupita- RV IV 5,7 (der Padapätha hat hier d-), oder es verlangt
nach einer anderen Erklarung: dtura- RV, ibhu- RV, dbhiika- AV k6nnen die Lokalpartikel
ä enthalten (Padapätha überall @-); aräti- ist MS I 5,1 : 67,5 unsichere Variante zu üblichem
drati-; asauca- ‘Unreinheit’ ist aus dsuci- unter Einfluß von sauca- abgeleitet (“Doppel-
vrddhi”); überdies ist es spät bezeugt (Sutra).

$ 11. Das Ergebnis lautet: @yavasa- ‘(Ort) ohne Weide’ enthält ein d-, das aus n und dem ur-
sprünglich anlautenden Laryngal von yavasa- kontrahiert ist. dyavasa- muß somit alt sein.
Die Beleglage ($8 1--3) läßt diese Annahme durchaus zu. Wie außerdem avest. yauuanha-
erweist, war das Grundwort mindestens im Urarischen bereits vorhanden”?. Das d-von
dyavasa- wurde deswegen nicht durch a- ersetzt, weil die Vokabel nicht mehr wirklich le-
bendig war?*. Auch im Padapätha des RV und der TS steht dyavasa-, nicht dyavasa- ; diese
letztere Form war wohl nicht in Gebrauch.

$ 12. Vielleicht fällt von ayavasa- auch Licht auf die vorurgriechische Lautgeschichte. yáva-
‘Gerste’ das Grundwort von yavasa-, ist ein Kognat von griechisch feıd ‘Getreide’. Es liegt
hier eine der bekannten Entsprechungen zwischen einem altindischen y- und einem grie-
chischen $- vor. Da nun das y- von altindisch yáva- auf ai- zurückgeführt werden muß, fin-
det die Ansicht, daß das $- von griech. feıa (und dann auch von $vyov usw.) lautgesetzlich
uridg. ai- fortsetzt®, neue Nahrung. Umgekehrt ist dann wohl anzunehmen, daß der uridg.
Anlaut i- im Griech. durch A- vertreten ist”.

Anmerkungen

| Die vier Fohlen des Masarsara, die drei des siegreichen Königs Ayavasa (erfreuten) mich’. Ähnlich

Renou, Ét. véd. et pan. 5 (1959) p. 7.

Zum Sachlichen s. Eggeling zu SB VI 2,3,2

Uber den besonderen Wert des Durva-Grases Eggeling zu SB VI 2 ,3,2. Vgl. auch AB VIII 8,6 ksatram
va etad osadhinam yad durva ‘der Herrscherstand unter den Pflanzen ist das Durva-Gras’.

Die Erwähnung von pasú- bringt es mit sich, daß die Handlung nicht nur dem Agni, sondern auch den
Tieren des Opferers zugute kommt (pasunam dhrtyai).

Bóhtlingk /Roth I (1855) hatten noch *Weideplatz, Futterplatz'; danach Grafmann in seinem Wörter-
buch.

Wilhelm Rau, Die vedischen Zitate im Vyakarana-Mahabhasya (Stuttgart 1985) p. 19. (= Akademie
der Wiss. und der Lit., Mainz; Abh. der Geistes- und Sozialwiss. Klasse, Jg. 1985, Nr. 4).

Vel. pasuvérdhana- RV ‘Gedeihen des Viehs'.

119

8 S, noch $ 4 (2. Absatz) zu KB XI 5. — Ein verwandter Gedanke z.B. RV V 53,13 rénan gavo né

10
li

yävase ‘sie sind vergnigt wie Rinder auf der Weide’, ebenso oder ähnlich I 91,13, IV 42,10, VIII 92,
12, X 25,1; pasür ná yávase in Vergleichen V 9,4, VI 2,9, VII 87,2.

Wackernagel/Debrunner, Altind. Gramm. II 1 p. 312, mit Nachtrag.

Nur V. zu Panini V 4,36 (vgl. oben $ 3).

suyavasad.-; süyavasin-; suyavasyü- (mit auffülligem Suffix, vielleicht zu suyávasa- nach einem Paar
wie avasyü- : avasá- gebildet).

Hier auch mit abweichendem Akzent (suyavasé). Zur Stelle s. Caland, AcOr 6 (1928) p. 107 (gegen
Eggeling, der hier eine besondere Bedeutung annimmt).

Weitere Prosastellen mit dem älteren suyavasa- : KausS XXIV 17; ‚ ApSS VII 28,7; VadhS bei Caland,
AcOr 6 (1928) p. 106 (bis). An der zuletzt genannten Stelle ist sizyavasa- übrigens maskulines Sub-
stantiv; die anderen Belegstellen für ‘gute Weide’ lassen Maskulinum oder Neutrum zu.

Die Bedeutung von kuyava- ist ibrigens undeutlich. Es steht in einem Opferspruch, in dem sich der
Opferer eine Reihe von — natürlich guten — Dingen wünscht: . .. pürnám ca me pürnátaram ca mé,
Ksitis ca me kuyavas ca mé, "nnam ca mé 'ksuc ca me... ‘mir r (soll zustande kommen) . . . Volles/
Volleres/ Nichtversiegen / kuyavah / Speise / Nichihaagern.: '. So TS; die Parallelen haben kuyavam :
MS Il 11,4: 142,1; KS XVIII 9: 272,4; KapKS XXVIII 9: 147,6; VS XVIII 10. Da etwas wie ‘Mift-
ernte’ hier nicht paßt, schlägt Aufrecht bei Böhtlingk (Sanskrit-Wörterbuch V p. 252 b) ansprechend
vor, 'küyavam (= aküyavam*) zu lesen und als ‘keine Mißernte’ zu verstehen. Die (maskuline oder
feminine) Pluralform der TS bleibt auch dann noch auffällig; vielleicht liegt hier ein Possessivum vor:
'(Jahre/ Zeiten o.ä.) ohne Mißernten’.

Diese Auffassung wird seit längerem vertreten; vgl. etwa Cowgill, in: W. Winter (ed.), Evidence for
laryngeals (London/Den Haag/Paris 1965) p. 165; Cowgill selbst ist freilich skeptisch.

Oldenberg, Noten zu RV IV 5,14. Hier (und zu V 12,4) auch über Versuche, das 4- von dsant- im
Kontext jeweils in d + á- zu zerlegen. Diese Lòsung ist fiir einen Teil der Belege ganz unwahrschein-
lich, naheliegend nirgends; auch der Padapatha gibt keine Handhabe.

Symbolae grammaticae in honorem Ioannis Rozwadowski I (Krakau 1927) p. 97 f.; Etudes indo-
européennes I (Krakau 1935) p. 30; zuletzt Indogermanische Grammatik II (Heidelberg 1968) p.
201.

Die Ersetzung scheint schon in der Sprache der RV-Dichter begonnen zu haben. Es ist wohl nicht
anzunehmen, daß RV asant- (mit @-) erst von den Redaktoren des RV stammt. Sie hätten es dann
eher überall durchgeführt; zumindest hätten sie kaum in verschiedenen Versen desselben Hymnus
(VII 104), ja im selben Pada (VII 104, 8d) äsart- neben dsant- geduldet.

So ist griech. vn-kepörjs (erwartet und jünger bezeugt: d-kepörjs) nach dem Muster von vnuepThe ge-
bildet, wo vn-< né- alt ist; s. R.S.P. Beekes, The development of the Proto-Indo-European laryngeals
in Greek (Den Haag/Paris 1969) p. 101 mit weiteren Beispielen.

Der Padapätha hat adeva-, aufer VI 49, 15.

Vgl. RV VII 104, 8d (s. oben Anm. 18) mit dsant- und dsant-; eine Nachahmung gerade dieser Stelle
ist aber wohl nicht weiter zu begründen. — Eva Tichy beabsichtigt, ddeva- in den MSS zu besprechen;
ich verdanke ihr wichtige Hinweise.

S. die Liste bei W. Wüst, Vergleichendes und etymologisches Wörterbuch des Alt-Indoarischen (Hei-
delberg 1935) p. 208; vgl. auch Bloomfield, Ved. var. II p. 230 f.

Der Wortausgang -asa- von urar. *hiauasa- ist auffállig; kann an ein Kompositum mit der Wurzel *së <
*seai ‘säen’ gedacht werden, also etwa ‘Gersten-Saat’?

Es wäre denkbar, daß im Arischen außer dyavasa- noch weitere, möglicherweise stärker verdunkelte
alte Privativbildungen unverändert mit a- < n-2- erhalten geblieben sind, doch ist bisher wohl kein
weiterer Fall nachgewiesen oder vorgeschlagen worden.

So zuerst wohl Sapir, Language 14 (1938) p. 271 f.; zustimmend Helmut Rix, Historische Gramma-
tik des Griechischen (Darmstadt 1976) p. 70.

Zu fordern wäre also der Befund: Keine anderweitigen Laryngalspuren bei den Entsprechungen ai.
y- : griech. h- (yah : ös); doch anderweitige Laryngalspuren bei den Entsprechungen ai. y- : griech.
$-. Anderweitige Laryngalspuren liegen tatsächlich vor bei yugd- : $uydv, nämlich u.a. ayunak RV,
trotz Cowgill (wie oben Anm. 15) p. 164; sowie bei yáva- : teid, nämlich dyavasa-, suydvasa- und viel-
leicht küyava-.

Abkürzungen (Texte, Literatur) in Anschluß an Johanna Narten, Die sigmatischen Aoriste im Veda
(Wiesbaden 1964). .

Brythonic gender reduction — the Cornish picture

Robert A. Fowkes (New York University)

We have only a vague notion of how genders were distributed in the lexicon of Indo-Euro-
pean, that is, the percentage of masculines, feminines, and neuters — for those are the
genders ordinarily assumed for Indo-European. There were probably three genders, at least.
Whether grammatical genders exceeded that number we cannot tell, and it is not easy to
imagine what additional genders would have been like if they had existed. (When certain
non-Indo-European languages are said to have many genders, the term is being used in a
different way).

In some groups of languages there has been a reduction of the number of genders. Hittite
had common and neuter, and the neuter differed from the common very little in most
cases. In Armenian the reduction was drastic and occurred prehistorically, for in all stages
of the language known to us there is no grammatical gender (it is hardly more felicitous to
say there is one gender, which amounts to none anyway).

In Celtic, as elsewhere, the neuter has been especially vulnerable. Old Irish had masculine,
feminine, and neuter gender. But early in the Middle Irish period (probably in the tenth
century) the neuter vanished, and nouns formerly neuter became masculine or feminine —
more frequently masculine. As Pedersen (1913: 66--67) pointed out, the incipient demise
of the neuter can be seen as early as Old Irish, with some originally neuter nouns vacillating
between neuter and masculine, and one or two between neuter and feminine. Thurneysen
(1946: 154) also calls attention to the partial elimination of gender distinctions in the
pronoun, the article, and the adjective long before the noun followed suit.

In Brythonic, according to a belief expressed by several scholars, the neuter was preserved
long enough for a number of Latin loanwords of neuter gender to be transmitted by Old
British to Old Irish, where they retained their neuter gender. Favorite examples are Old
Irish ór n. ‘gold’, from a British equivalent of Welsh aur (now masculine, like Modern Irish
ór) and Old Irish fin n. ‘wine’ (now masculine, like Welsh gwin, from Latin vınum n.)
(Pedersen 1913: 66). I am not persuaded that this guarantees that Old British still had a
neuter; there could have been re-assignment of the words to the neuter gender in Irish for
some reason now obscure. The British Latin words themselves could even have lost their
neuter gender.

At any rate, the Brythonic languages — Welsh, Cornish, and Breton — have only masculine
and feminine nouns. An attempt to see one solitary neuter in the Cornish word tra ‘thing’
(Nance 1938: 95) on the grounds that the word is treated as feminine in some grammatical
situations and as masculine in others, as is also true of the Breton equivalent fra, seems

122

more than a little far-fetched. Pedersen (1913: 67) regarded the treatment of Bret. tra
(feminine after the definite article: ann dra ‘the thing’, but masculine after the numerals
daou and tri: daou zra, tri zra ‘two things, three things’ with the specifically masculine
form of the numerals) as an indubitable trace of the ancient neuter. If, however, neuter is
to mean neither masculine nor feminine, tra seems rather to be both masculine and femi-
nine.

If the neuter gender was the first to be peculiarly vulnerable in Celtic (there seem to have
been neuters in Gaulish, Galatian, and Celtiberian, but their history is not known), its place
of maximum vulnerability was soon taken in the Brythonic languages by the feminine.
(Goidelic shows a similar development. In Old Irish, for example, the feminines were most
numerous and the neuters least. In Modern Irish the masculines are about one and a half
times as numerous as the feminines). Throughout the centuries the number of feminines
has declined, either through loss or through re-assignment as masculines. This is apparent in
lexical counts, which are, admittedly, not so important as contextual occurrence. For the
lexical distribution of gender, as seen in the relative percentages of masculines and femi-
nines in the dictionary, is obviously of less significance than the frequency with which
nouns of each gender are used in speech or writing. If some are used with decreasing fre-
quency until zero is reached and nobody remembers them, they may be said to have
vanished from the lexicon, whether by neglect, abandonment, expulsion, retreat, or ex-
piration — and all such words are no doubt largely figurative. In languages lacking a system
of writing, the loss is probably irreversible (although the retention of ancient vocabulary,
even of ‘obsolete’ words, may be part of the secret baggage of medicine men and the like,
who transmit them to successors, thus resembling on a small scale the use of extinct lan-
guages in the rituals of various organized religions.) In literate communities there is a
chance of retrieval and resuscitation. Modern German words like Abenteuer ‘adventure’,
Fehde ‘feud’, Halle ‘hall’, Hort ‘hoard’, Recke ‘heroic warrior’, Wonne ‘delight’, etc. had
once disappeared from the standard language but were revived and re-introduced into the
lexicon by authors of the eighteenth and nineteenth centuries. Sometimes the individual
author is identifiable and the date of the word’s revival known. Through their reading such
people were able to recapture part of the lost vocabulary of older periods. Sometimes the
revived form has a different gender from the earlier one (Abenteuer is neuter, but Middle
High German dventiure was feminine, e.g.). At times dialects also help in the process of
resuscitating once obsolete words, although the semantic direction taken by a word in the
dialects was often somewhat remote from the older meaning.

The lexicon is surely something greater than the dictionary, although dictionary makers
have often tried to do justice to the lexicon. In Celtic lexicography they have at times done
more than that: they have created their own words and flooded the dictionary with non-
existent vocables. Part of the task of modern Welsh lexicography, for instance, has been to
weed out spurious forms, ghost words, and other such extraneous matter. A quarter of a
century ago, when I attempted a preliminary study of gender redistribution in Celtic
(Fowkes 1957: 39—46), the best available Welsh dictionary was the twelfth edition of
Spurrell’s Welsh-English Dictionary (Carmarthen 1934), edited by J. Bodvan Anwyl with
the aid of a dozen foremost Welsh scholars. They succeeded rather well in ridding the
dictionary of most of the phantoms that had haunted it in the past. And they made admir-
able progress in distinguishing between obsolescent and obsolete words, and in marking
them. That task, however, was not completed to perfection, and the editor practically

123

admitted that in his preface. For, although a certain number of words could be designated
as indubitably obsolete, there was also a sizeable body of borderline words. Poets have
from time to time dipped into that store of words and exploited it; moreover other words
of biblical origin were not used in everday conversation but were well-known to hosts of
speakers. Some of this vocabulary was, with reluctance on the editor’s part, marked as not
belonging to twentieth-century Welsh or, to use a somewhat eccentric criterion, as being
words which the non-Welsh student of the language should not attempt to use in ordinary
prose. Those words were marked with a dagger: *. The absence of that sign should have
implied that a word was accepted as current Welsh of half a century ago. But many words
lacking that mark were not really current after all. Despite such difficulties I compiled
statistics on gender distribution based on Anwyl’s revision of Spurrell and arrived at the
following:

Masculines 7,892. Percentage of total vocabulary 67.41
Feminines 3,815. 7 t 2 32.59

Since the publication of that article a newer Welsh dictionary, has appeared (Evans-Thomas
1978), consulting editor of which was Professor Stephen J. Williams, an outstanding Welsh
scholar. First printed in 1958, the work has been through three revisions. The discrimina-
tion between obsolete and current words has been exercised with special care, and the
work is a more reliable one than all predecessors. (A monumental dictionary — a sort of
small-scale “Welsh Old English Dictionary’ has been appearing in fascicles since 1950 (Bevan).
By the time this article appears it will presumably have reached letter M).

Using Evans-Thomas (1978) and disregarding all obsolete words, I find the following
statistics:

Masculines 7339. Percentage of total vocabulary 71.32
Feminines 2999. E = 7 28.68.

This contrasts notably with the earlier figures of 67.4% for masculines and 32.59% for
feminines.

Exact computations are perhaps not as important as the direction of change. An increase in
the percentage of masculines over feminines is clear.

In Breton (Hemon 1973) the lexical figures are:

Masculines 12,711. Percentage of Vocabulary 56.6
Feminines 9,761. i B i 43.4

Breton thus exhibits less of a gap between masculines and feminines than does Welsh. But
the dominance of masculines is obvious. (Contextually, it is far higher).

Cornish shows closer agreement with Welsh than with Breton, at least lexically. A smaller
body of words is available in the dictionaries published (Nance 1938: 95, s.v. tra). Figures
obtained are:

Masculines 2,205. Percent? ol Vocabulary 74

Feminines — 794. S 26.

The picture alters considerably, however, when we count the Cornish nouns in their con-
textual occurrence in works of literature. I have done this for the four chief works preserved

in Middle Cornish. They are: Ordinale de Origine Mundi (The Beginning of the World),

124

Passio Domini Nostri Jhesu Christi (The Passion of our Lord Jesus Christ), Ordinale de
Resurrectione Domini Nostri Jhesu (The Resurrection of our Lord Jesus Christ), and
Beunans Meriasek (The Life of Saint Meriasek). The first three were edited and translated
by Edwin Norris (1869). The Beunans Meriasek was edited with a translation by Whitley
Stokes (1872).

These works are all plays in verse. There has been considerable debate regarding their date
of composition. The fourteenth century is accepted as the likeliest period by many, while
the Beunans Meriasek is regarded as somewhat later than the other three, while still in the
same century (Longsworth 1967: 7—8; Ellis 1974: 38). The gender statistics for these works
follow:

De origine mundi: masculine 1849 = 83.6%
feminine 363 = 16.4%

Passio masculine 1794 = 84.9%
feminine 319 = 15.1%

Resurrectio masculine 1517 = 84%
feminine 287 = 16%

Beunans masculine 1936 = 85%

feminine 349 = 15%

The Cornish lexical count, which yielded 70.12% masculines and 29.88% feminines, thus
gives an exaggerated impression of the percentage of feminines, which in actual literary
occurrence have at least 13% lower frequency than the dictionary might lead us to believe.

Three centuries later than the Middle Cornish works referred to above is Gwreans an Bys
(The Creation of the World), a miracle play written down by William Jordan of Helston in
1611, but apparently not composed by him (Stokes 1864). Whitley Stokes provided an
edition of this play, plus translation, in 1864. The statistics follow:

Masculines 1411 = 86%
Feminines 236 = 14%.

That is not an astounding change three centuries after the other works inspected, although
it does show a slight decrease in the percentage of feminine nouns. (One should suspect
that it is older than hitherto believed).

The next work used to derive similar statistics dates from the decade 1660—1670. It is a
story, “Jowan Chy-an-Horth, py an Try Foynt a Skyans” (‘John of Chyannor [Ram’s
House] or the Three Points of Wisdom’), written down, along with several other tales (that
have not survived), by Nicholas Boson on Newlyn, near Penzance, allegedly in order to
provide his children with a knowledge of Cornish (Nance 1949: 37—47). It is, unfortu-
nately, of insufficient length to have much statistical validity, containing only 222 nouns,
merely a fraction of those in the Middle and Late Cornish works studied. Of those nouns,
159 are masculine and 63 feminine, i.e. masculines amount to about 72% of the total,
feminines to 28%. Hence, approximately four centuries after the Middle Cornish works,
but only half a century after the Gwreans, if the date is right, there is an actual increase in
the percentage of feminines, although masculines still dominate conspicuously. But the
extent of this limited evidence dwindles even more when we note that those 159 masculine
attestations include only 50 different nouns, the most frequent, chy ‘house’, occurring 20
times; and the 63 feminines include just 21 different nouns, the most frequent, gwrek
‘wife’, occurring eleven times.

125

If there is dubious validity of such limited evidence, there may be still less justification for
considering writings in ‘Recent Cornish’, i.e., works composed after the language ceased to
be spoken. It might, then, not be legitimate to compare these relatively artificial products
with those written in contemporary Welsh or Breton. Since, however, there is a modern
Cornish version of the gospel of St. Mark (Smith 1936), I have, possible objections not-
withstanding, compared the gender distribution in that book in the three Brythonic lan-
guages. What emerges is the following:

Welsh: masculine 871 (67.595) feminine 420 (32.5%)
Cornish: masculine 1047 (78.4%) feminine 289 (21.63%)
Breton: masculine 946 (70.44%) feminine 397 (29.56%).

Here the Cornish figures, for whatever they are worth, do not show the extremes seen in
Middle Cornish, but the masculines continue to dominate, and the feminines are only 36%
of the number of the masculines. Furthermore, Cornish surpasses both Welsh and Breton in
the number of masculine nouns occurring, as well as in the percentage of masculines over
feminines. Now, although the translators of the gospel into Cornish were not using their
native language — nor, indeed, a living language — they cannot be suspected of manipulating
the vocabulary in order to keep the number of masculines high and the feminines low. In
fact, the one possibly conscious act on their part might be the avoidance of excessive
English loanwords, and that would keep the masculine count down. At any rate, by the
time of the demise, the Cornish vocabulary had reached a distribution of masculine and
feminine nouns such that the same passage, translated into all three Brythonic languages,
would require far few feminine nouns in Cornish than in the other two.

Why would the Cornish language, in all periods in which it is known, seem to shun the
feminine gender more than its cognate languages did (even though the latter also show
decline in the frequency of feminines, especially contextually)? Without falling into the
trap of teleological interpretation or of attributing to a language the ability to make a
choice, we can state that one of the embarrassing complexities of all Celtic languages is
the phenomenon of ‘initial mutation’. Even native speakers often find that phenomenon a
stumbling block. This sandhi-like process (but differing from sandhi in that the precipitating
causes are not usually extant, being prehistorical) is, for the most part, more of a burden
than a blessing. Feminine nouns induce far more initial mutation than do masculines. (In
Cornish there are one or two instances where masculines take a mutation where feminines
would not, but these are not of the same order of frequency by any means).

When words are borrowed from other languages, any Celtic language seems to assign nouns
of feminine gender in the source language to masculine gender in the recipient language.
And when the contributing language (especially English) has virtually no grammatical gen-
der, the chances are far greater for the word to become a masculine than a feminine in the
borrowing Celtic tongue. But all of these factors seem to apply to Celtic languages as a
whole. Why should Cornish reduce its feminine nouns at a greater rate than the others? The
process may be connected with the fact that Cornish is the one Brythonic language that has
become extinct. Was this excessive rejection of feminine nouns and the concomitant muta-
tional difficulties part of the process of rushing towards death?

There is considerable doubt as to when Cornish actually died. The story, long current, that
Dolly Pentreath of Mousehole, who died in 1777 at the age of 102 — allegedly — was the
last native speaker of the language is now treated with deserved skepticism. For there are

126

reports of native Cornish speakers long after her death (Norris 1869: 466; Jenner 1904:
21-23). Be that as it may, Cornish apparently died some time early in the nineteenth
century. As in the case of Manx (which, however, lasted until the 1970's), Cornish has its
presentday enthusiasts who read and write the language, and even speak it, to some extent.
I have heard a speech given in the Cornish language, and the sentence melody as well as
some of the sounds were very ‘English’, impressionistically. Talk of a Cornish revival can
only be called highly optimistic. But some people seem to write Cornish quite capably. A
mimeographed quarterly, An Lef Kernewek (The Cornish Voice), edited and largely
written by E.G.R. Hooper since 1951, has provided excellent Cornish prose and poetry.
(A sampling of its issues over the period 1960—1969 shows, incidentally, 76% masculine
nouns and 24% feminine). Despite such accomplishments of individual Cornish enthusiasts
(including some Englishmen and Americans as well as an occasional German), it has to be
admitted that the answer to ‘Is the Cornish language dead?’ must be in the affirmative.

In a number of admirable studies on the 'dying' Gaelic dialects in East Sutherland, espe-
cially that of the village of Embo, Nancy C. Dorian (1973: 413—438; 1976: 96—109) has
discussed the role played by initial mutations, chiefly lenition, not in the language death
itself but as symptomatic of the approaching end, perhaps. She analyzes the observance or
neglect by individual speakers in their treatment of lenition. There is great divergence be-
tween the speech of the oldest and the youngest generation. And, although there is a
certain determined retention of the initial mutations by all speakers, the younger speakers
may use a different (‘wrong’) mutation, or may use one where their elders have none, or
may use mutations interchangeably where the speech of the older folk allow only one, or
they may omit a theoretically mandatory mutation.

When failure is perceived in the operation of the system of initial mutations, such failure is
on the part of individuals or identifiable groups of individuals. Some short-comings are
regarded by other speakers (not committing the same offenses) as hilarious (Dorian 1976:
102); others may pass unnoticed by all but the most competent of fluent speakers. Some-
times fluent speakers themselves are ‘guilty’ of lapses. I have observed this in Wales. A
usually very capable speaker was heard to say several times, of children playing on the
beach at Aberystwyth, Maen nhw wrth ei fodd, which was intended to mean 'They're
having the time of their lives'. But faulty lenition (in this instance using it where none was
required) caused the utterance to mean “They’re having the time of his life’, and general
merriment ensued on the part of young and old alike that so prominent a personage and so
fluent a speaker could commit such a ‘blunder’.

This article points to a different aspect of the same process: the avoidance of situations in
which mutation (especially lenition) would be required, hence the predilection for mascu-
lines over feminines, since, as has been said, the latter call for more instances of lenition.!
That avoidance does not exactly constitute error, because the 'rules' of lenition are ob-
served in those cases requiring them. And the rejection of feminine nouns seems not
necessarily to be a conscious act, although it could be in a specific case. But we do not have
those cases, since nobody observed Cornish in the process of decay and death; or, if some-
body did, no reports were handed down to us. Both at the lexical and the contextual level,
however, the process seems to be an act of langue, rather than parole, if one may use those
terms in these times. [t is possible, of course, that in those instances in which a word
formerly feminine became masculine in Cornish, this occurred as a result of errors by indi-
viduals to start with, who used the words as if they were masculines. Then the error could

127

have been repeated so often that a generation arrived on the scene for whom they actually
were masculine nouns. In a way, the literary works from which the figures in the present
article are derived are the products of individual performance, yet the choice of words
must have been motivated by other considerations than the gender of nouns.

In shying away from the feminine gender, Cornish managed to transfer to the masculine
a number of nouns with suffixes that once marked feminine gender exclusively, or almost
so. For example, the Celtic suffix “kta, which has been exceedingly productive in Celtic
languages, yielding -acht/-echt in Irish, weth in Welsh, -ezh in Breton and -eth in Cornish,
was once almost a guarantee of feminine gender. (There is a chance that a small number of
such nouns were masculine in Celtic, possibly from *-aktu-) (Pedersen 1913: 32—33). In
Breton such nouns are still almost always feminine, and in Welsh about 92% of the nouns
in -weth are feminine (433 out of 473). Cornish sometimes has the old feminine gender,
sometimes vacillates between feminine and masculine, but often goes over to masculine
gender entirely. Welsh has esgobaeth feminine ‘bishopric’ (that esgob is a borrowing is not
relevant here), corresponding to Breton eskobiezh feminine (with a competing form
eskobelezh, also feminine), whereas Cornish epscobeth is masculine (although the first part
of the word is obviously more conservative than the corresponding Welsh and Breton
forms). Similarly, Corn. cudhyjygeth ‘contrition’ is masculine, while Breton keuzidigezh
(from which the Cornish word is possibly taken) is feminine.

Cornish has been aided and abetted in this shift away from the feminine gender of nouns
in -eth by the two-fold origin of -eth. It can be, like Welsh eth, from Celtic *-akta, (cf.
Dorian 1973: 413—438; 1976: 96—109), or it can be the equivalent of Welsh -edd (from
Celt. *-ria, which may have been feminine in origin but came to mark masculine nouns
too.) Eventually the masculines in -edd outnumbered the feminines in Welsh. Cornish
borrowed some of these, and additional support for -eth was provided as a masculine ter-
mination in Cornish. Thus, Corn. gwyryoneth *truth' is masculine, like Welsh gwirionedd,
and the feminine gender of Breton gwyrionezh is no doubt a retention of the early gender
(reinforced by -ezh from *2kta).

A number of loans of Norman origin, all apparently feminine to begin with, have entered
Cornish, perhaps at more than one period, and seem to be entirely masculine in their
Cornish guises, for example: benedyksyon, condysyon, conversasyon (‘manner of life’),
creasyon, dampnasyon, devosyon, examnasyon, nasyon, oblasyon, passyon, pensyon,
redempsyon, remosyon, salvasyon, etc. When there are Breton equivalents they are femi-
nine; Welsh ones vary, but feminines predominate.

A large number of other nouns which seem to be ultimately of French origin have mascu-
line gender where French has feminine. Some examples are: antyquyta, auctoryta, mar-
chondys, leouta (‘loyalty’), reouta (‘royalty’), cryjyans (‘credence’), aquytans, venjyans,
musyk, arsmetryk (‘arithmetic’), company, etc., all of which are masculine in Cornish. The
nouns in {y/ans show a suffix which accidentally coincides with a native one, -{yJans of
Celtic and even pre-Celtic origin. It is more productive in Welsh, apparently, than in
Cornish, cf. Welsh colliant ‘loss’, gogoniant ‘glory’, gwelliant ‘amendment, betterment’,
llwyddiant ‘success’, peiriant ‘engine’, etc., all of which are masculine. Cornish gordhyans
masculine ‘worship, honor’ is an instance of a native word with the same suffix, as is, most
likely, arghans masculine ‘money, silver’, Welsh arian, Middle Welsh ariant, Breton arc hant
masculine. It is conceivable that the native ending, associated as it was with the masculine

128

gender in Brythonic, provided the analogy for assigning that gender to the French loan-
words. It is difficult to decide whether these words came into Cornish directly from Nor-
man French or were transmitted via Middle English. If they were received from English at a
stage that knew no grammatical gender, it is not surprising that Cornish made them mascu-
line. Many of them seem to have been learned borrowings (ecclesiastical, especially) and in
such cases there could conceivably be restoration of the original French gender, but that
seems not to have occurred. (That the Breton gender matches the French is not surprising).

At times gender distinctions are useful in distinguishing homophones, cf., for example,
German das Tor ‘the gate’ vs. der Tor ‘the fool’; or der Leiter ‘leader’ vs. die Leiter ‘ladder’.
Many pairs of such homonyms could have been similarly differentiated in Cornish, yet that
advantage was often relinquished in favor of reducing the number of feminines. Corn. owr
‘gold’ and owr ‘hour’ are both masculine, although the second noun is from an earlier
feminine. There are many such pairs: ancar masculine ‘anchorite, anchor’; er masculine
‘defiance, eagle’; cost masculine 'cost, coast';porth masculine “doorway, port';etc., although
sometimes an old distinction is retained: toll masculine ‘hole’ vs. toll feminine ‘toll’.

If studies by Dorian (1973: 413—438; 1976: 96-109), Dressler (1977: 33—34), Szemeré-
nyi (1981: 281—310), Swadesh (1948: 226—235), and others have shown that certain
losses, confusions, and simplifications are symptomatic of imminent language death, it
must not be assumed that the symptoms cause that death. Nancy Dorian cogently pre-
sents her arguments that the failure of differential lenition in a dialect of East Sutherland
Gaelic is one type of change reasonably to be expected in 'decaying' and 'dying' Celtic
dialects. Other significant changes occur in morphology, too, for example the loss of the
genitive. The excessive favoring of masculine over feminine gender in Cornish is part of a
‘failure’ to observe many occurrences of lenition that would once have been mandatory.
It is pointless to speculate whether Cornish might have achieved a state of no grammatical
gender (like English, the overpowering neighbor, and possibly somehow because of that
neighbor?) had it survived. It is also mere speculation to guess that Welsh may be the next
Brythonic language to become extinct (since Breton has not ravaged its feminine nouns to
the same extent). It is also not to be taken as a sign of the approaching death of any lan-
guage when loss of grammatical gender occurs (English and Armenian refute that). Nor
would wholesale elimination of mutations in a Celtic language necessarily herald the death
of that language (although something similar to it is seen in the symptoms associated with
apparently imminent death in the Embo dialect studied by Nancy Dorian). In recent
decades a type of Welsh has come to be favored as something like a colloquial standard.
Called Cymraeg Byw (‘Living Welsh’) (Fowkes 1981: 107—118), it has simplified com-
plexities, eliminated some superfluous differentiations, and disentangled many intricacies.
It has not abolished the feminine gender, and it has not ruthlessly assailed the mutations.
Its proponents hope that Cymraeg Byw will help preserve the language. Time will tell.

Oswald Szemerényi (1981: 285—286), in discussing Wolfgang Dressler’s studies of certain
Breton dialects, many youthful speakers of which seem to disregard to a great extent the
mutations (a situation that might point to a rapid decline), makes the perspicacious ob-
servation that the prediction could easily prove false if the revived nationalistic enthusiasm
evinced by younger Bretons at the present time turns out to have a positive effect upon the
question of the language and to reverse the trend seeming to threaten the language with
extinction. Such a reversal obviously did not occur in Cornish. It may very well do so in
Welsh.

129

Dynargh, Henry Melgos! Mester yethow hag yedhyoneth, dysgador, scoler, densa, car. Dew
re-sowenno dheugh-why hag bedheugh-why lowenek y n bledhynnow a dhe.?

Notes

Lenition in Cornish: replacement of a voiceless stop by the homorganic voiced correspondent: p t K
become replaced by b d g; the voiced stops become corresponding voiced spirants v è y (the latter
realized as zero), etc.

Reluctantly rendered into Sawsnek thus: Greetings, Henry Hoenigswald! Master of languages and lin-
guistics, teacher, scholar, human being, friend. Godspeed and every happiness in the years to come!

References

1936. An Awayl herwyth Synt Mark, transl. by A.S.D. Smith assisted by R. Morton Nance (2nd edi-
tion 1960) (St. Ives: Lanham).

1951—. An Lef Kernewek (The Cornish Voice), edited by E.G. Retallack Hooper (Camborne:

Bevan, Gareth A., 1950—. Geiriadur Prifysgol Cymru: A Dictionary of the Welsh Language (Cardiff:
University of Wales Press).

Dorian, Nancy C., 1973. “Grammatical Change in a Dying Dialect”, Language 49: 413—438.

Dorian, Nancy C., 1972/1976. “Decay in Scottish Gaelic Language Death: The Differential Failure of
Lenition”, Word 28: 96-109.

Dressler, Wolfgang/R. Wodak-Leodolter, 1977. “Language Preservation and Language Death in Brittany”,
Linguistics 191: 33—34.

Ellis, Beresford, 1974. The Cornish Language and Its Literature (London: Routledge & Paul).

Evans, H.M./W.O. Thomas, 1978. Y Geiriadur Mawr: The Complete Welsh-English, English-Welsh
Dictionary (8th edition) (Swansea/Llandysul: Davies & Gomer).

Fowkes, Robert A., 1957. “Gender Redistribution in Celtic — A Preliminary Study”, in: Studies pre-
sented to Joshua Whatmough on his Sixtieth Birthday, edited by E. Pulgram (’s Gravenhage: Mouton),
39-46.

Fowkes, Robert A., 1981. “Some Characteristics of Modern Colloquial Welsh", in: Bono Homini
Donum: Essays in Historical Linguistics in Memory of J. Alexander Kerns 1, edited by Y.L. Arbeit-
man and A.R. Bomhard (Amsterdam: Benjamins), 107-118.

Hemon, Roparz, 1973. Nouveau dictionnaire Breton-Francais (Brest: Al Liamm).

Jenner, Henry, 1904. A Handbook of the Cornish Language (London: Nutt).

Longsworth, Robert, 1967. The Cornish Ordinalia (Cambridge, Mass.: Harvard University Press).

Nance, R. Morton, 1938. Cornish-English Dictionary (St. Ives: Lanham).

Nance, R. Morton, 1949. Cornish for All (St. Ives: Lanham).

Norris, Edwin, 1869. The Ancient Cornish Drama (2 vols.) (Oxford: University Press).

Pedersen, Holger, 1909-1913. Vergleichende Grammatik der keltischen Sprachen (Bd. 1 u. 2) (Göttin-
gen: Vandenhoeck & Ruprecht).

Stokes, Whitley, 1864. Gwreans an Bys (The Creation of the World), suppl. to Transactions of the
Philological Society (London: The Philological Society).

Stokes, Whitley, 1872. Beunans Meriasek (London: Trübner).

Swadesh, Morris, 1948. “Sociologic Notes on Obsolescent Languages”, HAL 14: 226-235.

Szemerenyi, Oswald, 1981. “Sprachverfall und Sprachtod besonders im Lichte indogermanischer Spra-
chen", in: Bono Homini Donum: Essays in Historical Linguistics in Memory of J. Alexander Kerns 1,
edited by Y.L. Arbeitman and A.R. Bomhard (Amsterdam: Benjamins), 281—310.

Thurneysen, Rudolf, 1946. A Grammar of Old irish, revised and enlarged, edited and translated by D.A.
Binchy and O. Bergin (Dublin: The Dublin Institute for Advanced Studies).

The Proto-Indo-European adpreps*
(Spatio-temporal auxiliaries)

Paul Friedrich (University of Chicago)

Universals

All ideas about relations are implicitly spatial, but some relational ideas are primarily so —
and of a concrete space at that. An irreducible minimum of such spatial categories is
integral to every morphology and syntax; it helps organize and constitute many fields of
meaning (of which it is a part); it is indispensable for interrelating the ideas in a text, and
for getting done the practical jobs of language: ‘Stand up!’ Like many if not most radical
or basic ideas, these spatial ones can be pointed at, gestured, or similarly represented
(Friedrich 1979: 452). The set is mapped below in Figure 1, as is consonant with its uni-
versality.

Figure 1: over, above

Universals
on to Z
back, behind
=
off from Z
D
fa
Lei
out of,
away from
in e
within 2
=
f =
front/opposite S
Lei
to, at, | underneath

toward

along

Also: between, among, with

* For their excellent comments on this paper I stand indebted to Phil Baldi, Rani Fedsin, Deborah
Friedrich, Paul Hopper, Bruce Mannheim, and James Redfield (the latter also tabulated the Homeric
Greek occurrences on the figure — about 1975).

132

Particulars

Universal spatial ideas are realized in the most diverse wavs. In the Tarascan language of
southwestern Mexico, for example, most such spatial relations are expressed by suffixes,
most of which have body parts as their primary meaning, be this as suffixes in long verbal
words, or as constituents of adverbial units: ču ‘bottom, buttocks, underneath’ (Friedrich
1971: 21 et passim); analogously, ‘inside’ and ‘chest, thorax’ are symbolized by another
suffix. Some Hittite and even Proto-Indo-European adverbs and prepositions decidedly
originated from words for body parts — for example, ‘heart’ and ‘inside, among’ (Ivanov
1973: 56—57); Wackernagel (1916: 182) thought that nominal bases could be assumed for
all such units, although some were probably deverbal (e.g., tr ‘to cross’ yielded trans
‘across’).

Depending on the language, spatial-directional relations may be expressed through case
endings, enclitics, particles, idioms, phrases, phonaesthesia, or the implications of syntactic
constructions (e.g., when they involve analogues between temporal, spatial, and transitivity
values). À most common device are the units known as adverbs or prepositions or others
that to a significant degree play the role of both these parts of speech — in short, the
“adpreps” (Bolinger 1971); the partly overlapping term, “adposition,” alludes only to
morphological slot.

Greek In Homeric Greek these adpreps number eighteen (the postpositions 6e, -dev, and
“pt are not included here). They are distributed throughout the texts and are frequent,
although some are more common than others; the first 250 lines of the Miad contain all
eighteen, but the rate, or range in frequency,.is from two that only occur once (èrép,
nepi), to èni, which occurs 22 times (as is consistent with its least-marked status). A full
list of the Homeric Greek adpreps is given in Table 1 in the order of their appearance in
Iliad 1; the first number after each adprep refers to the line in which it first occurs; the
second number refers to the number of times it appears in the first 250 lines of the Ziad;
these are the usual ‘proper’ prepositions.

Table 1

TPÒ 3 8 Tapa 34 6
én/é& 6 12 aut 37 5
bia 7 4 Kata 40 7
ovv/Evv 8 5 pera. 58 8
ava 10 9 eis/&s 71 5
éni 12 22 npös/norilnpori 84 9
év/évi/eivi/eiv 14 9 Dag 147 2
and 25 20 Dep 205 1
avri 31 3 mepi 230 1

What is the meaning, syntactic behavior, and status of these adpreps in Homeric Greek and,
to some extent, in Proto-Indo-European?

In Homeric Greek and other early Indo-European languages the adpreps may occur in
phrases that range from hapax legomena to frequent formulae. The adnominal adpreps, to
begin with, stand in two functional relations. First, they may disambiguate case categories:
for example, the genetive of source versus the ablatival genetive of motion away from.
Second, their own spatial meanings may be disambiguated by case — if an adprep can occur
with more than one case (Kurytowicz 1964: 176).

133

In fact, only eis goes only with one case, the accusative; otherwise, two adpreps go with
the genitive only (e.g., &), two go with the dative only (e.g., &vt), four go with the genitive
or the accusative (e.g. ŝi), and no less than eight go with any of the three oblique cases —
with differing semantic consequences, of course. In such multiple options, one of the co-
occurring cases reinforces the primary meaning of the adprep, whereas the others are 'ex-
ponents of a secondary function,’ or ‘specify the spatial relations expressed by the case
form’ (Kurylowicz 1964: 178). Some intimations of the overall semantic-syntactic com-
plexities are conveyed by Figure 2 of the subcategories of the Homeric Greek genetive,
which — in interaction with context — is often disambiguated by an adprep. (The so-called
‘improper prepositions’ such as äyxı all take the genitive, but, on the other hand, are not
preverbal).

Figure 2: possessive
The Homeric Genitive

. description
‘place names
descriptive name formulae
‚formal sets
psychic
partitive space/time
material
reciprocal
genealogical
source cause-material

ablative e
comparison

‘metonymic’ price
standard super/sub |

profit from,

deprive

lative P |
action“ ‘affective for

X from miss

touch

is lead, rule
allative affect against

action against

syntactic objective

subjective
genitive absolute

We find considerable freedom in order. The adnominal adpreps may be prefixed to sub-
stantives, either as mobile, detachable units, or as part of frozen compounds. They usually
precede the noun, but may follow (the reverse of the Sanskrit rule). They often occur
between adjective and noun, as in oàs émi vas, and this is a favored formulaic construc-
tion. They are never formally necessary, as in 70£' &porow Exwv, “bow [on his] shoulders
having (Wackernagel 1916: 197). But the rules and patterns for all these adprep-noun co-

134

occurrences, which are largely stylistic, have never been stated succinctly, and it would be
extremely difficult to do so.

As is clear already, cases also vary greatly in the degree to which they need an adprep. The
more sensitive the case is to context, the more rigid its binding to an adprep. Other early
dialects have stricter rules. In the Rigveda the fourteen ‘genuine prepositions’ are ‘almost
entirely restricted to employment with the accusative, locative, and ablative... and were
originally connected with the accusative, locative only' (Macdonnell 1971: 210). Latin is
relatively rigid when it comes to these and other case relations (Baldi 1977).

The degree of overt, linear attachment to the verb runs a wide gamut from (1) direct pre-
fixation to a bound verbal form, to (2) direct prefixation to a free verbal form, to (3)
detached from the form but directly preceding it, to (4) separated by one word only, to (5)
separated by two or more words, to (6) postposed directly after the verb, to (7) separated
by two or more words from the preceding verb (in both (6) and (7) the adprep gets the
stress), to (8), where the adprep and the verbal form are in adjacent lines. When not direct-
ly prefixed, the meaning of the adprep and, hence, of the expression in which it occurs (or
simply of the rest of the expression), may remain unaffected; that is, AdprepVerb=Adprep
... Verb, as in of Kara fovs nadtov. In other cases, the discontinuity may decrease the
bonding of the adprep with its verb and increase its bonding with a noun. Often a new
ambiguity is created; thus, tepuupéw means ‘to strip off,’ but zept ppévas iuepoc aipet
seems to include this meaning and also something like ‘seized all around,’ that is, ‘desire
seized his heart all around.’ In other cases there are subtle semantic differences reminiscent
of English ‘to cut the bread up, to cut up the bread.’ Other things being equal, the farther
the adprep is from its verb the greater is the likelihood of changes in meaning of the sort
just adumbrated.

When primarily bound to a verb, the adprep, in what is its second main function, disambi-
guates the aspectual subcategories of the verbal expression. By ‘aspect’ is meant the tem-
poral values inherent in the activity, event, or state itself (Jakobson 1957: 493; Holt 1943),
particularly as regards the features of duration, instantaneity, inception, completion, and
boundedness. In Homeric Greek this aspect was of at least four main kinds, some more
powerful than others but never enough so that, logically, one can include the other. The
subcategories are given here in their order of power and of textual frequency.

First, an adprep can denote or at least adumbrate the relative instantaneity or punctuality,
as against the durativity, of an event or act ‘to sail along (mapa) the shore.’ Second, an adprep
can denote or imply completion (or lack of totality): ‘to cut off” (äro-raurw), ‘to cut’
(rauvw). Third, several adpreps denote direction toward a goal or simply direction (telic or
determinative aspect): mpdc, for example. Finally, several adpreps denote realization or the
full result of an action. In all four of the meanings just given the adpreps are in some sense
bounding the action, as opposed to not bounding it (this category of bounding in aspect
theory is best argued by Jakobson). The strong case for boundedness as the deepest se-
mantic feature is, of course, more or less congruous with the pragmatic or discourse-
oriented claim that these are perfectivizing mechanisms to provide for sequencing, for
closing off one event before another can begin to be narrated.

The aspect value of adpreps is often, perhaps most typically, actuated with verbs of mo-
tion. Moreover, the subcategories interact with each other and two of them are often
realized simultaneously in one form; thus, the telic often interacts with the durative/

135

punctual or the completive/noncompletive. The adprep ovr is particularly interesting
because it usually has the telic meaning of ‘the reunion of objects’ (Brunel 1939: 107), as
in ‘to tie together, to come together.’ But this may combine with the aspect of the verbal
root (Aktionsart) to denote completion, as in ‘I shatter completely’ (ovvayvumn). Latin per
is also used this way. Here are the first six adpreps with examples of some of their aspectual
functions (Table 2).

Table 2

appi intensifies durative value of augirérouai durative
‘to be busy at’

ava Good ‘I snatch up’ punctual

amo ùtorduvew ‘I cut off? completion

bud daneprw ‘I destroy utterly, sack’ realization

eis eioapıkavrw ‘I arrive at a destination’ telic

eK Ek 'yeAdw ‘I laugh out’ inceptive

Among the variables governing the relations between adprep and verb one must include the
following: (a) the (degree of) boundedness of the verb; (b) the degree of attachment
between adprep and verb; (c) the degree of (allowable) distance between adprep and verb;
(d) the freedom of the adprep to occur after the verb; (e) the number of adpreps that have
the freedom of distribution to occur with a given verb; (f) the number of verbs that are free
to occur with a given adprep; (g) the changes in meaning — including subtle nuance — con-
nected with adprep association; (h) the number and identity of adpreps that can occur with
a given verb at the same time (which runs up to three); (i) the bondedness of the adprep
with the nouns and their case meanings in a phrase or other construction; (j) the interac-
tion between the adprep and its noun, the adprep and its verb, and so forth. Actually, an
adequate understanding of how adpreps work would call for delicate measures and ex-
haustive lexicography of the relative affinity between specific adpreps and whatever they
were in construction with, both in general and in the sense of specific verbs — granted the
high degree of idiomaticity and unpredictability and, indeed, the low feasibility of the
entire project. (Cunliffe and similar lexicographers are excellent on the scope and, to some
extent, the frequency of these co-occurrences, but there remains the problem of com-
binatory potential; compare Friedrich (1971), where an attempt is made to give the com-
binatory potential of all Tarascan verbal roots with all suffixes.)

The complexities and realities of the adprep systems are obscured in many ways by the
conventional conceptualizations and nomenclatures. Adpreps, for example, actually occur
more often unattached, and this is the unmarked order. Rather than saying that the un-
attached adpreps are in tmesis, which would be correct for Attic Greek or Old Norse, we
should use a special term for the marked, attached adpreps — perhaps ‘in synthesis.” Simi-
larly, the traditional nomenclature of adverb, preverb, preposition, and postposition has
— except where these are used for special situations — obscured the more basic reality that
the adprep is always typically in construction with both the verb and the nominal object,
as most or at least many Homerologists and Indo-Europeanists have recognized (Meillet,
Benveniste, Ivanov, Chantraine). The adpreps are underlying verbs — specifically auxiliaries —
and can, with the verb in ellipsis, function as surface predicates, as in Gv 'Oó6vooeUc
TONUUNTIS.

136

Syntactic Trees Here are two typical examples of the sort of construction in question:

à»à OTpaTOV Wyxero xa delow

*the arrows [k5Àa] of the god [9€iow] went [coxero] against [avà] the army [orparov]
aupi yàp cououow (áAero Eüpov

‘about [ài] his shoulders [@poow] he threw [BaXero] his sword [£ipor]

I assume that the meaning of the adpreps is to be found, not only in the forms, but in the
way they combine with other units in phrases and clauses in stanzas, episodes, similes and
yet other discourse-length spans. How are the sliding scales of governance to be represented
syntactically in terms of trees, bracketings or some other explicit, if necessarily simplified
form?

There are four main kinds of construction to be represented. In the first the adprep is
primarily in construction with the noun and both are subordinate to a verb (Kurytowicz’s
“first type”). In the second, equally important type the adprep is bound mainly to the
verb, on which it is dependent; for example, [NP [Adp V]] (assuming a NV order). Third,
the adprep is in construction with the verb in the verb phrase, and, indirectly, with a noun
phrase subordinate to that verb phrase. Last, the adprep is simultaneously in construction
with both the verb (phrase) and the noun (phrase). This is the least-marked proto-syntagm,
the syntactic archetype. The probiem of its looseness will be taken up below.

NP VP

Meaning

The primary meaning of the Homeric Greek adpreps was spatial: ovv ‘with’ (that is, side by
side), amo ‘off, away,’ and so forth — spatial meanings such as are laid out on Figure 1
above. But the meanings, as shown above, are often temporal as well; as Chantraine says
(1963: 82), they are ‘little invariable words, originally autonomous, which make space-time
relations more precise.’ These space-time markers are qualitative, never quantitative (whereas
the noun (phrases) in an aspectual unit often are quantitative); but the way the adpreps
work to express at least a dozen subcategories of aspect is congruous with the multifarious
way aspect is quantified through adverbs or nominal adjuncts. Incidentally, the same
technical term seems to have been coined independently about the same time by Ivanov
(1973: 60: sluzhébnoje postrdnstvennoje imja) and Friedrich and Redfield (1973-74:
locative auxiliary); however, ‘spatio-aspectual’ now appears more precise than ‘locative,’
and ‘adprep’ is not only conveniently short but conveys more information than ‘auxiliary’.

The semantics of the adpreps entails other theoretical dimensions. For one thing, there are
important semantic relations between aspect and case. The latter denotes what are, at least
implicitly, spatio-temporal ideas; for example, away from a space (ablative), into a space

137

(accusative). Such meanings are variously congruous with aspectual ones: the accusative
case with the telic aspect, the nominative with the durative, and so forth. The Homeric
adpreps, finally, are interstitial between grammatical (‘pure relational”) elements, such as
affixes — since they are always potential prefixes — and, on the other hand, the more con-
cretely referential words in the language (Sapir 1921: 107). This structural interstitiality
or hybrid status may be one reason they have been negiected or misclassified as a set in
many general books on Proto-Indo-European.

Adpreps in Contexts

The aspectual value of the adpreps should not be neglected, but it also should not be ex-
aggerated: these meanings are far less powerful than the spatial ones. Let us take a fairly
typical Homeric Greek adprep, kará, which has a strong cognate in Hittite kata, and a
Proto-Indo-European etymon. Meanings that are primarily or exclusively aspectual define
only two out of eleven or, by another count, 2/18 of the categorial total, and its total
aspectual role in texts would also probably come to about one fifth of the total; the
following Figure 3 summarizes some of these patterns. Readers desiring the full, actual
contexts need only look up in Cunliffe each of the forms cited in this figure and go from
there to the Homeric Greek loci.

Figure 3

ee intensive
Aspect nn.
completive karakéyw

Ai over
; d Lauda
Verbal—___ Locative& — iin Karà yaiav
ba under
down

source kar 'obpavoo

Genitive E goal
place

KATA YUÑS

distributed karà orparóv
Nominal

toward
at
in, into
through
against
Accusative across = allomotive
abreast of
reflexive

according to Karà poipav

138
Proto-Indo-European

Homeric Greek is a well-marked stage in the transition from Proto-Indo-European and Pre-
Greek to the typologically much changed structures of Attic and Hellenistic Greek (Meillet
1934). The complex drift had several interconnected components that are integral to the
adprep question. First, there was case syncretism, the ablative fusing with the genetive, and
so forth. With the loss of the ablative, locative, and instrumental cases, the functions and
frequencies of the adpreps increased, and the adpreps ‘proper’ tended to be replaced by
‘improper prepositions,’ and by new adverbial forms derived from nouns. Second, the
adpreps fused with specific verbs in the process of ‘univerbation,’ so that eventually an
adprep was never free of its verb. Third, word order became much more fixed; adpreps, for
example, came regularly to precede their nouns. Fourth, the adpreps came to agree more
regularly with the more limited number of cases, particularly the accusative (just as certain
conjunctions became associated more rigidly with certain moods through rules of taxis).

From these well-documented, interdependent, multivariate drifts (Sapir 1921: VII), we
can extrapolate backwards toward Proto-Indo-European, and posit adpreps that were
relatively autonomous and unattached and, concretely, about equally in construction with
nouns and verbs (Wackernagel 1916: 165, 178 et passim). A full panoply of six of the eight
cases were variously and loosely in construction with the adpreps, as were various aspectual
forms of the verbs. These hypotheses, which seem sound on the basis of Greek alone, are
strongly supported by other early dialects, notably Vedic and Hittite. In Vedic, “the place
of the preposition is free ... in that, in principle, it can appear before or after the noun”
(Renou 1968: 169—70). Similarly, ‘‘adverbial prepositions follow, but also precede, their
case (Macdonnell 1971: 208). Of the cognate situation in Hittite, Pedersen wrote (1938:
151): “Es hat im Hettitischen insofern ein altertiimliches Aussehen, als die meisten Pro-
verbia noch dem Verbum gegenüber sehr selbständig sind, also eigentlich Adverbia sind, bei
denen es im Einzelfall oft unmöglich zu entscheiden ist, ob sie sich einem vorhergehenden
Kasus also Postposition anschließen oder als Praeverbia zum folgenden Verbum gehören.”
See also the more recent statement by Otten and Soucek (1969: 86—88).

In sum, adprep rules were relatively probabilistic. Adpreps occurred about equally before
and after nouns in Proto-Indo-European (with markedness inconclusive). Vis-a-vis verbs the
least marked position was probably the preverbal one, although they were free to occur
postverbally. The most marked position was probably clause initial. My hypothesis is that
most of the adpreps could and usually did stand in construction with both nouns and
verbs, but that a few were mainly adnominal while a few others mainly modified verbs;
some, at any given point, were in various stages of syntactic change and differentially
bonded and with differential freedoms of distribution. I think we have to assume that the
meaning of the adpreps would -- depending on the type of bonding — fluctuate consider-
ably in degree of abstraction and transparency. Finally, it would in principle be possible to
construct a markedness diagram for all 23 Proto-Indo-European adpreps (granted the
prodigious problems of coding, multivariate analysis, and lexical detail needed for the ex-
position). We are left, for now, with the archetypal formula given above as the most general
and accurate description.

Indo-Europeanists tend to attribute phonological and morphological contrasts back to the
reconstructed parent language (for example, the voiced aspirates, the eight cases). Then we
derive the forms of the descendent languages by loss and simplification — or better, what I

139

would call symmetricization (interacting with complication through analogy, and so forth).
It is interesting that — by what looks like the obverse of the same assumptions and proce-
dures — the tendency is to posit a relatively simple, loose, and flexible proto-syntax (e.g.,
Meillet 1964: VII — for a contrasting argument see Lehmann 1974). But there is a real
difference between the phonological and syntactic situations and the attendant problems
of reconstruction. A phonology at an earlier stage (e.g., Old English) is not generally more
complicated than at a later stage (e.g., Modern Texas English), whereas the syntax of many
of our texts in early Indo-European is demonstrably looser in the sense, for example, that
optional rules are more powerful and marked forms more frequent. The syntactic looseness
in question might well be expected in the extremely poetic texts of such languages as
Homeric Greek and the Rigveda (Wackernagel 1968: 88 et passim thinks such looseness is
one diagnostic feature of poetic language). But it also seems likely that the language of
ordinary discourse (conversation, narrative, and so forth) was more poetic in Proto-Indo-
European times. This point need not be over-played, however: many of our early texts are
not particularly poetic (e.g., in Hittite, Old Persian, Old Latin) and we also depend on their
relatively prosaic language for inferring the meaning and syntactic function of, inter alia,
the adpreps.

As for their meaning, the Proto-Indo-European adpreps were originally primarily spatial
(Wackernagel 1916: 210), although there was great dialectal variation (Wackernagel 1916:
154). They were also aspectual, and many of them contributed a perfective sort of meaning
(Wackernagel 1916: 181, also Chantraine 1903: 40), whereas others denoted various other
categories and subcategories of aspect. These ancillary aspect functions resemble each other
in our most critical texts: Homeric Greek, the Rigveda, Hittite, and Old Irish; the texts in
Old Russian and other languages are also often priceless.

The actual phonemic form of the adpreps has been or can be reconstructed with as much
certainty as can any other part of the early vocabulary or morphology, and is thoroughly
documented in the major dictionaries by Frisk, Pokorny, and the like. The forms are set
forth on the table below (I don't know that they have ever been brought together before as
a set). The layout runs: the proto-form, then the main spatial meanings, then one or two
hypothetical aspectual meanings, then examples from three or four stocks, and then the
total number of stocks with cognates for the form in question. Incidentally, one reason
that the inference of the aspectual value of the Proto-Indo-European adpreps must remain
hypothetical is that it is so hard to reconstruct syntagmata with unbound adpreps. The
meticulous work by Kuhn, Wackernagel, Schmidt, and the like, has never to my knowledge
included a syntagm with an adprep — to say nothing of a free adprep. This reflects one
limitation of the comparative method. In many cases there are several reconstructable
forms with overlapping or roughly the same meanings (e.g., many of Pokorny's entries),
and it is difficult to work out a hierarchy. Lehmann (1974: 232) lists eight such cognate
forms for Proto-Indo-European “pro; in some cases these sets can be ordered in terms of
inflection and/or ablaut (e.g., pro-, pre-, pero-; (Meillet 1964: 350)).

140

Table 3

*ad, ‘to at, telic

towards, by’

*ambhi, ‘on both perfective,

sides, around’ durative

*an, anu, ‘up, ingressive,

upwards, along’ telic
egressive

*anti, ‘facing,

opposite’

*apo, ‘away, off inceptive

*dis-a, ‘through’ realized

*do, ‘to’ telic

*eks, eghs, ‘out of? inceptive,

*en, ‘in, into’ liminal

*enter, ‘into, liminal

between’

*epi, opi, ‘near telic,

on’ inceptive

*et, ‘away, beyond’

*knta, ‘down’ completive

*medhi, ‘middle, telic(?)

among’

*ni, ‘downwards’

*para, ‘beside, durative

along’

*peri, ‘across, durative(?)

around’

*pos, ‘behind,

right by’

*pro, pero,

‘before, ahead of’

*proti, ‘to,

against’

*s, ‘with, side telic

by side”

*tras, ‘across’

Lt. ad, Olr. ad, Go. at

al, Osc. ampt, Alb. mbi,
Olr. imb-, Ger. um; Gugoo, Lith.
abu, Toch. A/B

apa, Av. ana, Go. ana, Arm.
am-, ham-

avri, Lt. ante , Hitt. anda

amo, Sk. dpa, Go. uf, Old Cornish a
ôá, Lt. dis, Go. dis-, Alb. tsh-

-6e, OE to, Olr. do, Russ. do

é£, Olr. ess-, OCS iz , OPr. esse

€», Arm. i, ON i, Av. ni-, Toch. A/B
Y- yr-

évrepov, Arm. ander-k; ON idrar,

Lt. inter

èri, RV api, Hitt. apa, Irish Ériu, ‘Ireland’
(*epi-uerio)

eri, Lt. et, Toch. A atas, OCS otù,
Sk. áti

kará, Hitt. kata, Olr. cet-

uerá, Alb. mjet, Go.mip , OCS me?da,
Sk. mádhyu

Sk. nî, Av. ni, OHG nidar, OCS niza
(maybe Gk. ves)

napa, Lt. prae, por-, Go. faur,

Sk. purd

nepi, Sk. pári, Go. fair, Lith. per,
Russ. pere-

moc, OCS poze, Lith. pas, Alb. pas,
Lt. post

100, Go. fra, Sk. pra, Olr. ro,

Hitt. paraa

mpos, npott, RV prati, Lt. pretium,
OCS pretiva

our, EUv, OCS cz, Lith. sù, Myc. Gk.
ku-su

RV tiras, Av. taro, Lt. trans, Olr. tar

4 stocks

7 stocks

5 stocks

4 stocks

6 stocks
4 stocks
3 stocks
6 stocks
10 stocks

9 stocks
9 stocks

5 stocks

3 stocks
6 stocks

4(5) stocks
5 stocks
6 stocks
7 stocks
7 stocks
4 stocks
3 stocks

3 stocks

141

*uperi, ‘over, out’ brép, Av.upairi, i ver, ON yjir, 7 stocks
Olr. for

*upo, telic bro, RV úpa, Go. uf, Myc. Gk. upo,

‘underneath, Lt. sub

from behind’

Naturalness and Nuance

That aspect is a natural ancillary value of the spatial adpreps is supported in other ways. In
one part of the Indo-European area, we find that the English prepositions enter into peri-
phrastic constructions to yield a panoply of phrasal compounds with a rich semantic
texture that is partly aspectual. In a very different part of the Indo-European area, both
geographically and typologically, we find the cognate Russian adpreps functioning ad-
nominally, and centrally, of course, in the aspect morphology. The aspect of the adpreps in
Russian and English could be illustrated just as well by Hindi or Welsh or any other Indo-
European language — or any language with such components (and I know of none that
entirely lacks them). The burden of proof, in short, rests with the proponents of the null
hypothesis: that the Proto-Indo-European adpreps did not have aspectual functions.

The aspectual information in a complex verbal form is never exclusively binary, even in
Russian, but always involves a plurality of aspect values in a network of aspect values.
These values vary enormously from one adprep to another: some are 'exclusively' (i.e.,
almost exclusively) binarily aspectual in a specific case (e.g., the s- of spet’, the perfective
aspect of ‘to sing’). In other cases various degrees, clines, and intersections are involved, or
it is a matter of nuance, and no Russian adprep lacks such nuances. In English the adprep
compounds similarly involve all manner of nuance, with meanings that range from fairly
simple lexical ones to others so subtle that only a native speaker can command them, that
is, use them without semantic accent (e.g., ‘to rip, to rip up, to rip through, to rip off,’
etc.). The same appears to be true of all natural languages and the assumption of such
natural nuance seems realistic and fruitful for Proto-Indo-European studies.

Conclusion

Ultimately, the universal scheme mapped at the outset is adequately labelled by the full set
of Proto-Indo-European forms on the final table above, forms which were used to com-
municate about physical space and interpersonal relations, and to handle communication-
ally indispensable aspect categories in the verbal system. I have assumed that Proto-Indo-
European was a natural language, which means at least two things. First, it was like any
known language with respect to syntactic form, pragmatic needs, semantic values, and so
forth. Second, naturalness also means that the actual form-meaning units, here the adpreps,
were adumbrated and complicated by language-specific nuances that are harder, but not
totally impossible to recapture; such nuances of aspect might, felicitously, constitute the
subject of a sequel analysis in the next festschrift for Henry Hoenigswald.

142
References

Baldi, Philip, 1977. “Morphosyntax and the Latin genitive", Folia Linguistica 11: 107—122.

Benveniste, Emile, 1968. Problémes de linguistique générale (Paris: Gallimard).

Bolinger, Dwight, 1971. The Phrasal Verb in English (Cambridge: Harvard University Press).

Brunel, J., 1939. L aspect verbal et l'emploi des préverbes en grec, particulièrement en attique (Paris:
Klincksieck).

Chantraine, P., 1963. Grammaire homérique 2. (Paris: Klincksieck).

Comrie, Bernard, 1976. Aspect. (Cambridge: Cambridge University Press).

Cunliffe, Richard J., 1963. A Lexicon of the Homeric Dialect. (Norman: University of Oklahoma Press).

Fraser, Bruce, 1974. “Review of Bolinger”, Language 50: 56-75.

Friedrich, Johannes, 1960. Hethitisches Elementarbuch (Heidelberg: Winter).

Friedrich, Paul, 1971. The Tarascan Suffixes of Locative Space. Meaning and Morphotactics. (Bloom-
ington: University of Indiana Press).

Friedrich, Paul, 1974. “On Aspect Theory and Homeric Aspect", JJAL Memoir 28. Vol. 40 Number 4,
Part 2.

Friedrich, Paul, 1975. Proto-Indo-European Syntax. JIES Monograph No. 1, Butte, Montana.

Friedrich, Paul, 1979. Language, Context, and the Imagination. (Stanford: Stanford University Press).

Friedrich, Paul and James Redfield, 1972—75. Lectures on Syntax. Chicago.

Frisk, Hjalmar, 1960. Griechisches etymologisches Wörterbuch. Vol. 1 (Heidelberg: Winter).

Frisk, Hjalmar, 1970. Griechisches etymologisches Wörterbuch. Vol. 2 (Heidelberg: Winter).

Hoenigswald, Henry, 1960. Language Change and Linguistic Reconstruction. (Chicago: University of
Chicago Press).

Hopper, Paul J., 1982. (ed.) Tense Aspect. Between Semantics and Pragmatics. (Amsterdam: Benjamins).

Holt, J., 1943. “Etude d’aspect.” Acta Jutlandica 15.2.

Ivanov, V.V., 1973. “The typological development of the Slavic and Indo-European prepositions and
postpositions”. In: Structural-Typological Investigations of the Grammar of the Slavic Languages.
Moscow (in Russian).

Jakobson, Roman, 1957. Shifters, Verbal Categories, and the Russian Verb. (Cambridge, Mass.: Harvard
University Press).

Kurytowicz, Jerzy, 1964. The Inflectional Categories of Indo-European. (Heidelberg: Winter).

Lehmann, Winfred P., 1974. Proto-Indo-European Syntax. (Austin: University of Texas Press).

Macdonnell, Arthur A., 1971. A Vedic Grammar for Students. (London: Oxford University Press).

Meillet, Antoine, 1934. Aperçu d'une histoire de la langue grecque. Part 3.

Meillet, Antoine, 1934. Introduction à l'étude comparative des langues indo-européennes. (University,
Alabama: University of Alabama Press).

Otten, Heinrich and Vladimir Soucek, 1969. Ein althethittisches Ritual für das Königspaar. (Wiesbaden:
Harrassowitz).

Pedersen, Holger, 1938. Hittitisch und die anderen indogermanischen Sprachen. (Kóbenhavn: Munks-
gaard).

Pharr, Clyde, 1963. Homeric Greek. (Norman: University of Oklahoma Press).

Pokorny, Julius, 1959. Indogermanisches etymologisches Worterbuch. Vol. I. (Bern/Miinchen: Francke).

Renou, Louis, 1952. Grammaire de la langue védique (Paris: Les Langues du Monde).

Sapir, Edward, 1921. Language. (New York: Harcourt, Brace).

Schmitt, Rüdiger, 1967. Dichtung und Dichtersprache in indogermanischer Zeit. (Wiesbaden: Harrasso-
witz).

Schmitt, Rüdiger, 1968. (ed.) Indogermanische Dichtersprache. (Darmstadt: Wissenschaftliche Buchge-
sellschaft).

Wackernagel, Jacob, 1920. Vorlesungen tiber Syntax mit besonderer Beriicksichtigung von Griechisch,
Lateinisch und Deutsch. Vol. 1. (Basel: Birkhauser).

Wackernagel, Jacob, 1943. “Indogermanische Dichtersprache’’, pp. 83--101. In: Schmitt, Rüdiger.

Regular contact dissimilation

Hans Henrich Hock (University of Illinois)

1. Ever since Osthoff and Brugmann’s (1878) ‘neogrammarian manifesto’, historical lin-
guists commonly have distinguished between ‘regular’ sound changes (such as assimilation,
weakening, and non-dissimilatory loss) and ‘sporadic’ ones (such as dissimilation, meta-
thesis, and haplology). Osthoff and Brugmann explained this distinction by considering
regular change to be ‘mechanical’, only phonetically conditioned, but sporadic sound
change to be conditioned by ‘psychological’ factors. These latter were identified by Brug-
mann (1909) as the "horror aequi(voci)", the avoidance of repetition. As time progressed,
also distant assimilation — except for vowel assimilation in neighboring syllables — was
found to be sporadic. Apparently for this reason, Paul (1920) invoked a broader concept to
distinguish between regular and sporadic change, namely that the latter is conditioned by
the same factors which are responsible for speech errors or slips of the tongue: incorrect
monitoring of the sequencing of speech events and the difficulties involved with speech
production which are exploited in tongue twisters.

2. It is the merit of Henry Hoenigswald (1964) to have pointed out a fundamental diffi-
culty with this postulated clear distinction between regular and sporadic changes: a number
of the 'shining examples of regularity' belong to the allegedly sporadic class of changes,
such as the Greek and Sanskrit aspirate dissimilations covered by Grassmann's Law.

Hoenigswald’s tentative solution to this difficulty lies in considering the distinction ‘regular’
vs. ‘sporadic’ sound change invalid: “... rather than say that there are maverick processes
in which sporadicity and distant conditioning are alarmingly combined, it could be, that
distant conditioning produces the appearance of sporadicity, simply by leaving so much
more scope for morpheme boundaries to intervene, and hence for analogy to operate.”

3. In this latter conclusion, Hoenigswald's view probably will not find many adherents.
Thus, when Lat. fragrare appears as both flagrare and fraglare in Vulgar Latin, or cribrum
as criblum and cribum (Schopf 1919: 80—83 etc.), it is difficult to attribute these varying
outcomes to analogy. Moreover, Togeby (1964) comes to the conclusion that of the many
Romance dissimilations noted in Posner 1961, only three can be considered regular, all
others being sporadic. Even of these three “regular” processes, two are of doubtful regular-
ity: The dissimilation of r to/ in the context V... V ... [t stop] r, while amazingly regu-
lar, fails to apply in aratru : Span. arado. And the explanation of the development of
meretrice to menetrice (beside meletrice) seems somewhat far-fetched. The situation is not
better for the dissimilation r > / in structures like arbore : Sp. árbol, robore : roble. Among

144

the exceptions we find not only the possible ‘cultismo’ martyre : mártir, but also priore :
prior, sartore : sastre.

At the same time, however, the fact remains that tbere are exemplars of the allegedly
sporadic class of sound changes which are as regular as any of the typically regular changes.
Hoenigswald's conclusion that the distinction between 'regular' and 'sporadic' changes is
not as clear-cut or absolute as the neogrammarians would have it, therefore remains as an
important challenge to historical linguists.

4. One possible way of meeting that challenge would be to establish that typically sporadic
changes can apply in a regular fashion only under certain well-defined, special circum-
stances. This would permit retention of a basic distinction between regular and sporadic
changes, while at the same time providing a principled account for the apparent exceptions
to this distinction.

In this paper I will try to show that there is one area where it is possible to establish such
special circumstances. This is the area of contact dissimilation, including the dissimilation
of vowels in neighboring syllables (cf. the regularity of vowel assimilation in the same en-
vironment), but excluding dissimilatory loss.!

While this is a quite limited sub-area of the larger field of ‘sporadic’ changes, it is hoped
that the findings of this paper, combined with Hoenigswald's challenge, will stimulate
further research on the question of whether a principled account can be given for the ap-
parently exceptional regularity of otherwise irregular processes.

5. In 1963, Benediktsson proposed that certain difficulties with the u-umlaut ofa in Eastern
dialects of Old Norse be explained as the result, not of two stages of umlaut (as had gener-
ally, but implausibly been assumed), but of a secondary “undoing” of u-umlaut in just
those cases where the conditioning environment u had remained; cf. the chronology in (1).
The explanation for this development, in his view, lies in the fact that before retained u
there was a lack of contrast between original @ and umlauted 9. (On the other hand, after
apocope, 4 and 9 had come to contrast elsewhere; cf. all vs. oll in (1).) The resulting archi-
phoneme then could be realized either as a (as in the Eastern dialects), or as 9 (as in Old
Icelandic), or (conceivably) as something in between.

(1) pre-ON allaz alli allum(—) ‘all’ (sg. N m., f., plD)
umlaut ollù ollum(—)
apocope all oll
Olcel. all oll ollum
OENorse all oli allum

6. Outside Germanicist/Scandinavianist cirles, Benediktsson's proposal does not seem to
have found the attention it deserves. Consequently, when Schane wrote his famous 1971
paper in defence of the notion (though not a separate level) of the phoneme, he was ap-
parently unaware of Benediktsson's paper. At the same time, he quoted and discussed a
number of examples which are strikingly similar to the Old Norse “reversal” of u-umlaut.

Thus, in the history of French, vowel nasalization first was introduced before all nasals and
later was eliminated where the conditioning nasal had remained and where thus there was
no contrast between nasalized and non-nasalized vowels; cf. (2). Similarly, in Nupe, palat-

145

alization and labiovelarization are in the process of being eliminated in just those contexts
where there is no contrast between palatalized (etc.) and non-palatalized (etc.) consonants.
But before a, where there is a contrast, no change can be observed; cf. (3). Schane finds a
similar situation in the case of Rumanian dispalatalization. Other parallel cases of dispalat-
alization can be added from many of the Slavic dialects (e.g. Ukrainian), from Lithuanian
and Latvian, as well as from Old Irish, where however the process is limited to non-dentals;
cf. (4).

(2) pre-OF bon bons
nasalization bón bone
final N-deletion bo
denasalization bono

attested forms — bó bon(a)

(3) Nupe egi ~ egi ‘child’
egWu ~ egu ‘mud’
vs. eg’ a oe e ,
egWa “hand” # ega ‘stranger
(4) Olr. endim [knap¥] : cnamf(a)i [knapi] ‘bone(s)’
(vs. sil [sul] : suili [sulYi] ‘eye(s)’)

From the facts which he had examined, Schane concluded that (a) changes of this sort are
limited to those positions where the relevant contrast is neutralized; (b) these develop-
ments involve the loss of a feature of “secondary” articulation (such as vowel nasalization
or consonant palatalization); (c) under such conditions, changes of this sort are expected to
occur.

7. Schane thus apparently independently has come to conclusions quite similar to Bene-
diktsson’s. However, in so far as they differ or go beyond the latter, they must be viewed
with scepticism.

Thus the fact that Russian and many other languages have no dispalatalization before front
vowel (in spite of a lack of contrast) and that Old Irish dispalatalization does not affect
dentals, shows that changes of this sort are not as automatic as Schane claimed.

Moreover, in cases like the dialectal Old Norse change of 9 to a it is difficult to conceive of
the process as one which eliminates a secondary feature of articulation. The case is similar
for Russian, where an early change of o to e after palatalized segments, leading to a neutral-
ization of o and e, is followed by a dissimilatory backing of accented e to o in part of this
neutralized environment; cf. (5). Likewise, in Modern Icelandic n goes to d before n in an
environment in which # and d do not contrast:? and under similar conditions, / becomes d
before /; cf. (6). But again, it is difficult to view these changes as eliminations of secondary
articulation features.
(5) © Slav. o>e / [t pa.] —
*znajomù > znajemù ‘we know”
(ii) Ru. é€>6 / [+ pal] [— pal.]
ber’ em > ber>’öm (“berem”) ‘we will take’

146

(6) Icel. (i) n>d / VV n°
einn > [eidn] ‘one’

(i) 1>d / l

all > [adi] ‘all’

8. Examples like (6) further show that the noted developments in neutralized environment
are not just simple "'reversals" of earlier changes. For the clusters nn and I of Old Icelandic
were not necessarily derived from earlier dn or dl. In fact, nn never reflects an earlier dn.

Rather, I want to argue that these changes are cases of regular contact dissimilation and
that it is moreover precisely the condition of contrast neutralization between the input and
the outcome of the change, common to all of the above examples, which is, if not a neces-
sary, then at least a highly conducive condition for regular contact dissimilation in general.

A cursory examination of the data dealt with in examples (1) through (6) will show that all
of the changes result in segments which are more dissimilar to the neighboring, conditioning
segments than the earlier, unchanged sounds. Moreover, all of these developments are
regular (with the possible exception of the Nupe change (3), which is still in progress).
Finally, in all changes there is no contrast between input and outcome in the environment
for the change. At the same time, outside these neutralized environments the changes do
not take place.

Further examples of this sort, also regular, can be found elsewhere. Thus Pedersen (1909:
158 and 162, 170) notes that nn and mm change to dn and bm in Cornish. As far as I can
tell, however, there are no contrasting structures with original dn and bm. Dieterich (1908:
81) observes a change of // to /t in certain Modern Greek dialects. Original /t, however, had
become rt in these and other Greek dialects; cf. Dieterich (1908: 65).*

In Sanskrit there are constraints of varying strictness against retroflex s before r or another
retroflex s (cf. Wackernagel 1896: 232—234 with Debrunner's Nachtráge 130—131). One
of these seems to have resulted from regular dissimilation, namely that in (7) below, where
$2 s/. r. (Itis trme, r is not retroflex, but alveolar. Phonologically, however, retroflex
and alveolar behave alike in triggering n-retroflexion; cf. Hock 1979.) Here as elsewhere,
the outcome of the change and its input did not contrast, due to the fact that in the en-
vironment after 7, u, original s had gone to * > s by RUKI. There is, to be sure, one excep-
tion, namely ajusran. But this form can be explained as due to leveling.

(7) () PIE tfr)isr-es ‘three (fem.)’
Plir. tifras (cf. Av. tiro)
Skt. fisras

(ii) similarly usar- ‘dawn : usras (sg. G)
reduplic. sarisrpa- < *sarisrpa- (from srp- ‘creep’)
reduplic. sisrate < *si-sr-ate (from sr- ‘run’)°
cf. reduplic. si-sic-ur (from sic- ‘pour’)

Also the change of xs to ks in most of the Germanic dialects (as in *uxs- : OE ox, ON oxe,
NHG Ochse [ks]) probably belongs here, with regular dissimilation of fricative x to k
before fricatival s. True, Modern German has apparently unshifted forms like machst [xs]
‘you (sg.) do’. However, these go back to forms like OHG mahhost, with a vowel inter-
vening between the x and the s. The modern [xs] can therefore be explained as due to the

147

fact that the change of xs to ks had taken place before the loss of intervening vowel in
forms like mahhost. (Alternatively, the [x] of machst can be explained as reintroduced by
generalization from forms like OHG mahhöt, NHG macht ‘does’.) Similar arguments are
possible for forms like OE @cs ‘axe’ which seem to offer original k before s, suggesting that
the contrast between k and x had not been neutralized before s. As the alternative Old
English form æces and especially Go. agizi show,æcs goes back to *ak(“Jiz/s-, with a
vowel intervening between X and the sibilant. The change xs > ks therefore may have
predated the loss of that intervening vowel, at a time when X and x were not as yet in con-
trast before s.

9. The data so far presented, then, strongly suggest that at least in the area of contact dis-
similation, it is possible to discern a special condition which permits regular dissimilation,
namely a lack of contrast between input and outcome in the environment of the change.

Evidently, in order to test whether this conclusion can stand as a universal condition for
regular contact dissimilation, it would be necessary to examine all instances of regular
contact dissimilation which ever occurred and to see whether they meet this condition.
Negative results would then require that either we place certain limitations on this ‘uni-
versal’ condition or (if they are too common) that we add the condition to the large pile of
‘junked’ universals.

As it turns out, it is quite difficult to find genuine instances of regular dissimilation. Most
textbooks, even articles and monographs exclusively dealing with dissimilation (such as
Grammont 1895 or Schwyzer 1933) do not distinguish between regular and sporadic dis-
similation. (The majority of listed cases, however, turn out to be sporadic.) Moreover, in
many cases the history of a given language is not sufficiently well known to permit an
unambiguous interpretation of apparently dissimilatory developments.

There are, to be sure, certain types of changes which are commonly enough attested as
regular processes to make it appear that they can be used as evidence against our hypothe-
sis. One of these is the common development of 7/ to KI (as in Lat. *potlo- 7 poc(u)lum;
vet(ujlu- > PRom. *veklu-; Lith. eglé ‘fir’ < *edle, cf. OPr. addle, Po. jodła). This change,
with dental t becoming velar k before dental /, has been called dissimilatory (cf. e.g. Stang
1966: 107). And, though regular, this change has taken place in languages with a con-
trasting X before the / (cf. e.g. Lat. oc(uJlu- - PRom. *oklu-). However, it is questionable
whether # > kl is simply a dissimilation. For there seems to be a cross-linguistic difficulty
with sequences of non-sibilant dental * /, a difficulty which may be resolved in several
different ways (beside the change # > Kl), such as anaptyxis (as in coll. Engl. [æ8aliyt]
‘athlete’), metathesis (as in OSpan. dadlenos > daldenos, tidle > tilde), or resyllabification
(as in Mod. Span. atlas [at$las] vs. the ordinary syllabification of stop + / observed in
coplas [ko$plas]). It is this cross-linguistic difficulty, rather than dissimilation, which seems
to be responsible for the change tl > kl.

Another common process which has often been called dissimilatory (cf. e.g. Rohlfs 1950:
57-58, 65) is that of the type kt, pt >xt, ft, with change of stop to fricative before stop.
At least in the case of Greek there is a chance that this change operated at a time when
structures like autos had changed into aftos (via avtos), thus producing a contrast between
f and p before f. However, fricativizations of this sort are also found before (fricatival)
sibilants, as well as before other consonants. (Cf. also section 12 below.) They simply are
instances of weakening. (Cf. Hock 1976.)

148

10. More specific potential counterevidence is found in the developments of Lat. *genmen
> germen, *kanmen > carmen. For the change of n to r before m looks like dissimilation;
the change is regular (in so far as both items in which it could have applied show it); and
there is a contrast of n with r before m (cf. e.g. arma, sermo). Note however that the
number of examples (two) is so small that it becomes almost meaningless to speak of
regularity. Moreover, the change has also been attributed to dissimilation with the later
following n (cf. Leumann 1926:231), i.e. as a distant assimilation.

A more ample set of potential counterexamples consists of the apparently dissimilatory
development of nasal n to non-nasal r after nasal m in Spanish (as in hominem 7 (h)omne
> (h)omre > hombre). For as humerum > (h)omro > hombro shows, there was a con-
trasting sequence mr. Note however that the development of hominem to hombre is not
simply the result of sound change. For hombre replaces earlier (hjueme. Even the latter,
however, probably is not directly inherited; for as dominum > dueño shows, inherited
-omin- develops to -ueñ- just like original -omn- (as in somnum > sueño). Both hombre and
(hjueme therefore no doubt represent borrowings from (partly vernacularized) Medieval
Latin. Grammont (1895: 138—139) therefore probably is correct in seeing in forms like
hom(b)re nativizations of Med. Lat. omne (with vernacularizing syncope) into a language
which had lost the sequence mn and which as a consequence used mr as the phonetically
closest native substitute.

11. More serious is the evidence of Old English and Old Norse, where we find apparently
regular dissimilatory developments of the type (8), in spite of the fact that the input to the
change contrasted with its output. (In the following I draw heavily on data from Luick
1914 and Heusler 1931.)
(8) (i) OE cysip > (*)cysb > cyst ‘thou choosest’
later OE piefb > pieft ‘theft’
{ge)sihb > gesiht ‘sight’
cf. wast ‘thou knowest’
heeft ‘prisoner’
pohte ‘thought’

(ii) ON *reiside > *reisbe > reiste ‘erected’
cf. veist ‘thou knowest’

As it turns out, however, the developments in question are not always regular. Thus later
Middle English has forms like sucp/zicp for OE sihp ‘sees’, with dissimilation of the first,
rather than the second fricative. Moreover, the changes in (8) are associated with other pro-
cesses which it would be hard to motivate as dissimilatory, cf. e.g. (9):

(9) ( OE *aubmodipo > *éadmedp(—) > éadmett- ‘humility’
(ii) ON *sakibò > (*)sekb > sext ‘guilt’
*diupibo > *dypb > dyft ‘depth’
*wohslide > *@kslde > gkslte (“gxlte”) ‘grew’
*halßiðö > helfd [vd] beside dial. helfp/t half

Though the details of these developments are too complex to be traced in this paper, it
may be suggested that the changes in (8) — as well as in ME sucp/zicp, or in (9) — are not
simple dissimilations. Rather, they seem to have been responses to certain structural con-

149

straints brought about by earlier changes: Grimm's Law, in combination with other devel-
opments, had led to a situation in which fricative-initial affixes such as the preterit marker
-dé alternated with stop-initial forms, as in (*nazjan- ‘save’ :) *nazi-dé beside the rarer type
(*bankjan- “think’ :) *bäax-te. Synchronically, this alternation would seem to be best ac-
counted for by a rule changing fricatives to stops after fricatives. (This is of course virtually
the opposite of the historical change, by which voiceless stops had changed to fricatives,
except after obstruent.) To this was later added the change of xs to ks. From both of these
structural facts it was then possible to reinterpret or infer a more general constraint against
double fricatives or (for Old Norse) a canonical form fricative plus stop for all (voiceless)
obstruent sequences (except those of the type stop + sibilant); etc. In some cases (such as
OE gesiht vs. ME such/zich) the constraint might be implemented in conflicting ways.
Finally, for forms like ON óksite, even more complicated generalizations seem to be in-
volved.

12. The potentially most damaging counterevidence against the hypothesis of this paper
comes from the history of Greek: As is well known, not only did kt, pt become xt, ft in
Modern Greek, the latter merging with the outcome of aut, eut > aft, eft; also khth, phth
changed to xt, ft; and sth, skh, and to some extent sph became st, sk, and sp. As is also
known, outside these clusters, the old voiceless aspirates kh, th, ph became the fricatives
x, 9, f. It is the relationship between these developments which makes for potential chal-
lenges to the hypothesis of this paper.

Now, as noted in section 9, the change of kt, pt to xt, ft probably is not dissimilatory, but
involves weakening. In fact, early Naxian NaHsios (with héta), and the development of ks,
ps to ss in Attic Vase inscriptions and elsewhere suggests weakening of stops to fricatives
also before fricative, at least in certain varieties of ancient Greek. The Modern Greek of
Southern Italy shows a similar fricativization even before n (Rohlfs 1950: 57, 64), and
traces of such a change can be observed elsewhere in Modern Greek, as in Lesb. d0na,
Comm. Gk. dxna, Pont. áfna/áxna, ultimately derived from C1.Gk. atmizò (Hatzidakis
1892: 95). Weakening thus may once have had a very broad scope in the history of Greek.
If outside of Southern Italy the effects of weakening appear only sporadically before nasal,
this may reflect the diglossic influence of the archaizing standard language, just as in the
case of the usual reflex sf for ancient sph. (The latter is almost certain to be explained in
this fashion. For sp, sometimes beside hypercorrect sf for old sp, is found all along the
periphery of Modern Greek, from Pontic in the East (Hatzidakis 1892: 162) via Cypriot
and other Southern dialects (Dieterich 1908: 79—80) to Italian Greek in the West (Rohlfs
1950: 65, 72—73).)

There is however a potential problem in the fact that khth, phth merged with the outcome
of kt, pt, i.e. with xt, ft, and that sth, skh (sph) became, and merged with, st, sk (sp). If
now it is assumed that these changes took place as in (10), with dissimilation in the frica-
tive clusters created by the general fricativization of the old aspirates, then we might be
confronted with a case of regular dissimilation, even though the input and the outcome of
the change were in contrast:

150

(10) okto khthes esthai steira
Weakening oxto
Fricativ. x0es esGe
Dissimil. xtes este
Attested oxto xtes este stira
‘eight’ ‘yester- ‘be’ ‘sterile’
day’

However, the exact chronology — and nature — of the changes is difficult to establish. At
least one author, Dieterich (1898: 96—101), has on the basis of inscriptional evidence
argued for a very different scenario, which can be restated as in (11).’ In this scenario, the
deaspiration after fricative would be typologically comparable to the absence of aspiration
in English and German voiceless stops after tautosyllabic fricative.

(11) oktö khthes esthai steira
Deasp. I” estai
Weakening oxto xthes
Deasp. II” xtes
Attested oxto xtes este stira

In this scenario, two processes of Deaspiration are invoked, so as to account for the se-
quencing of events as stated by Dieterich, who finds evidence for sth > st to be completed
by the third century B.C., but for kt/khth > xt only from the fifth century A.D. However,
Schwyzer (1939: 210—211) cites ‘é€xx7av (with unaspirated, non-fricatival £) from a fourth-
century B.C. inscription from Cos. If this could be taken to indicate an early date for kt/
khth > xt, then the two Deaspirations could be collapsed and could be ordered after
Weakening.?

Given this possible alternative interpretation and the uncertainty of interpretation of the
epigraphical evidence, it seems unnecessary to see in the above Greek developments
counterevidence to the hypothesis of this paper.

There is, however, evidence that just as in Old English and Old Norse, the patterns and
alternations created by the above changes (cf. e.g. Mod. Gk. apfo) to > af to) made possi-
ble a reinterpretation according to which there is a general constraint against (voiceless)
stop clusters or (voiceless) fricative clusters. And just as in Old English and Old Norse, this
reinterpretation seems to have spawned off a number of developments, some of them dis-
simulatory, some not; some regular, some not; some in conformity with the conditions for
regular dissimilation postulated in this paper, some not. Thus xs, fs have become Ks, ps in
almost all of Modern Greek, with the exception of Italian Greek (Rohlfs 1950: 63—64,
74-75); ss has changed to és in a number of Greek dialects, some of them having ts from
other sources (Dieterich 1908: 80), others apparently not (Pernot 1907: 286—291). The
most extensive documentation of these changes (including such developments as r@ > rt,
rx > rk, yO > yt, yr > kr) is found in Pernot (1907) who also points out that in terms of
geographical distribution, many of these changes must be considerably later than kt/khth >
xt and sth > st. Moreover, many of the changes which he cites are incomplete or irregular.
For these and other reasons, his conclusion that all Modern Greek consonant clusters tend
to become clusters of voiceless fricative + voiceless stop is only tentative.

151

13. While the evidence discussed in the preceding two sections thus does not provide
cogent counterarguments to the hypothesis of this paper, it may require the following
slight modification: The reinterpretation of certain phonological structures as exhibiting a
contraint against particular segment combinations may lead to response changes which may
include regular contact dissimilation, even in cases where the input and outcome of the
change are in contrast with each other. At the same time, however, they may include other
changes which cannot be easily viewed as dissimilatory.

Except for this possible modification of our hypothesis, however, the discussion in sections
9-12 has failed to turn up any cogent counterarguments, initial appearances to the con-
trary notwithstanding. Combined with the positive evidence presented in sections 5—8, this
fact would seem to strengthen the case for the claim that regular contact dissimilation re-
quires the absence of contrast between input and outcome in the environment of the
change.

14. The most striking evidence for the validity of this hypothesis seems to come from the
development of geminate sibilants in the prehistory of Sanskrit. For here the regularity vs.
irregularity of a contact dissimilation seems to be directly attributable to the absence or
presence of contrast, respectively, between input and outcome of the change. (The data, to
be sure, are prehistoric, postulated intermediate stages in the development from Proto-
Indo-European to Sanskrit. However, I believe to have shown elsewhere (Hock 1974, 1975)
that they constitute plausible and necessary steps in this development.)

Disregarding an archaic degemination, by which double sibilant ss became s and which
survives only in a few relics, we can note that Proto-Indo-Iranian geminate sibilants under-
went regular contact dissimilation in the case of 3% (> pre-Skt. retrofl. ss) and ss, but failed
to do so in the case of ss; cf. (12). The reason for this different behavior can be sought in
the fact that ss was in contrast with ts (cf. (12), while there was no comparable contrast
between fs and ss or cs and ss. While the latter sibilants therefore were free to undergo
regular contact dissimilation, in the case of ss, the dissimilation took place only sporadical-
ly (cf. apassu vs. vatsyati).

(12) PIE wiks wiksu -ske/o- opessu wessyeti weytsyeti

PIIr.I wiss wissu -s$a-

PIIr.IE visé visu -$a- apassu vassyati vaitsyati

pre-Skt. viss vissu

Dissim. vits vitsu — -c$a- — -—--- vatsyati

Attested vit viksu . -ccha- apassu vatsyati vetsyati
‘clan’ ‘clan’ pres. ‘active’ desid.> desid. >
(sg.N) (plL) suffix (pl.L) fut. of fut. of

vas- vid-

15. Given the evidence presented in this paper, it seems safe to conclude that Hoenigs-
wald’s challenge concerning the neogrammarian distinction between regular and sporadic
sound change can be met in the area of contact dissimilation: With the possible exception
noted in section 13, it can be stated that contact dissimilation can become regular only in
those cases where the input to the process and its outcome are not in contrast in the en-
vironment for the change.

152
Notes

I also exclude diphthongization, a process which could be conceived of as dissimilatory, but for
which many other explanations are possible; cf. e.g. the discussion and references in Wanner 1975.
Modern Icelandic, to be sure, offers a few forms like buddna [büdna] ‘of purses’. However, these
seem to be recent borrowings. Moreover, if inherited, the [dn] of these forms can only reflect earlier
ddn, not simple dn.

A. systematic exception consists of forms with suffixed definite article, such as ey-nni (sg.D f.) ‘to
the island’. These can be explained either in terms of relative chronology or as analogical: Forms like
ævi-nni (sg.D.f.) ‘to the life” suggest that -nni at one point was preceded by short i, which would have
blocked nn > dn. Alternatively, the -nni of ey-nni can be viewed as analogical to the regular -nni of
forms like ævi-nni.

Also initial | may change to /f. This may either be attributed to external sandhi (Schwyzer 1933) or
to the wide-spread gemination of initial consonants in Southern Greek (cf. Dieterich 1908: 61—62).
Forms like sisarti ‘runs’, with s > s before non-neighboring r may either be explained as generaliza-
tions from sisrate etc. or as instances of the less regular constraints against s before distant s or r.
Perhaps the Germanic change of mn to vn, as in *stemnò : Go. stibna ‘voice’, ON stefn ‘summons’,
belongs here, too. However, some of the details are difficult to establish and opinions on the inter-
pretation vary. (Thus in 1897: 383, Brugmann considered the change regular, while in 1903: 115, he
stated that the situation is 'unklar', and Prokosch (1938:87) considers the change sporadic.) While
the change appears to be quite regular for Gothic and Old Norse, there is room for the suspicion that
mn in some cases became mm, rather than vn; cf. e.g. OHG stamm, OE stem(m) 'stem', ON stemma
‘to stop, stem’ vs. OS stamn, OE stefn, ON stafn ‘prow, stem’ (cf. e.g. Brugmann 1903: 115).

I have retained my weakening explanation for the preconsonantal stops, rather than subscribing to
the quite speculative and complex hypothesis of Dieterich according to which the change progressed
as follows: pt > pht > phth > pfth > pft > ft. As a consequence I have to invoke a second deaspira-
tion process.

The difficulties with interpreting the epigraphic evidence can be illustrated by the fact that from the
second century B.C. onward we find both forms like karandınevos (Chios) and like the ‘exxrav
quoted above. Schwyzer (ibid.), taking forms with 70 at their face value, sees in the former an inter-
mediate stage pth. However, xxr suggests much earlier structures of the type kt or xt. (Dieterich
takes them to represent kxt.) Perhaps the contradictory evidence is best explained as reflecting neu-
tralization in the vernacular language of the earlier aspirate and unaspirated stops of these clusters,
combined with attempts in the more archaizing educated language to retain the distinction in ques-
tion. (Cf. the discussion in Sturtevant (1940: 83—85) on the retention of aspirate pronunciation in
archaizing, educated Byzantine Greek, long after the vernacular change of these aspirates into frica-
tives.)

References

Benediktsson, Hreinn, 1963. "Some aspects of Nordic umlaut and breaking", Language 39: 409—431.

Brugmann, Karl, 1897. Grundriß der vergleichenden Grammatik der indogermanischen Sprachen 1
(Straßburg: Trübner).

Brugmann, Karl, 1903. Kurze vergleichende Grammatik der indogermanischen Sprachen (Straßburg:
Trübner).

Brugmann, Karl, 1909. “Das Wesen der lautlichen Dissimilation’”. Abhandlungen der Sächsischen Gesell-
schaft der Wissenschaften, Philosoph.-Historische Klasse 27: 139-178.

Dieterich, Karl, 1898. Untersuchungen zur Geschichte der griechischen Sprache . . . (Leipzig: Teubner).

Dieterich, Karl, 1908. Sprache und Volksüberlieferungen der südlichen Sporaden (Schriften der Balkan-
kommission 3:2.)

Grammont, Maurice, 1895. La dissimilation consonantique dans les langues indo-européennes (Dijon:
Darantiere).

Hatzidakis, G.N., 1892. Einleitung in die neugriechische Grammatik (Leipzig: Breitkopf & Härtel).

153

Heusler, Andreas, 1931. Altisländisches Elementarbuch, 3rd edition (Heidelberg: Winter).

Hock, Hans Henrich, 1974. “Historical change and synchronic structure: the case of the Sanskrit root
nouns”, in: Toward tomorrow's linguistics (Washington: Georgetown University Press) 329—342.

Hock, Hans Henrich, 1975. “Substratum influence on (Rig-Vedic) Sanskrit?”, SLS 5 2: 76-125.

Hock, Hans Henrich, 1976. “Final weakening and related phenomena”, in: 1975 Mid-America Con-
ference Papers (Lawrence: University of Kansas) 219—259.

Hock, Hans Henrich, 1979. “Retroflexion rules in Sanskrit”, SALA 1: 48-62.

Hoenigswald, Henry, 1964. “Graduality, sporadicity, and the minor sound change processes”, Phonetica
11: 200-215.

Leumann, Manu, 1926. Lateinische Laut- und Formenlehre (München: Beck).

Luick, K., 1914. Historische Grammatik der englischen Sprache, reprinted 1964 (Oxford: Blackwell).

Osthoff, Hermann/Karl Brugmann, 1878. Vorwort: Morphologische Untersuchungen auf dem Gebiete
der indogermanischen Sprachen 1 (Leipzig: Hirzel).

Paul, Hermann, 1920. Prinzipien der Sprachgeschichte (Sth edition, reprinted 1975) (Tübingen: Nie-
meyer).

Pedersen, Holger, 1909. Vergleichende Grammatik der keltischen Sprachen 1 (Göttingen: Vandenhoeck
& Ruprecht).

Pernot, Hubert, 1907. Études de linguistique néo-hellénique 1. ((Paris:] Published by the author).

Posner, Rebecca R., 1961. Consonantal dissimilation in the Romance languages (Oxford: Blackwell).

Prokosch, Eduard, 1938. A comparative Germanic grammar (Baltimore: Linguistic Society of America).

Rohlfs, Gerhard, 1950. Historische Grammatik der unteritalienischen Grazitàt (= Sitzungsberichte der
Bayrischen Akademie der Wissenschaften, Philosophisch-Historische Klasse 1949: 4).

Schane, Sandord A., 1971. “The phoneme revisited”, Language 47: 503-521.

Schopf, Ernst, 1919. Die konsonantischen Fernwirkungen (Göttingen: Vandenhoeck & Ruprecht).

Schwyzer, Eduard, 1933. “Dissimilatorische Geminatenauflösung als Folge von Übersteigerung”, KZ 61:
222-252.

Schwyzer, Eduard, 1939. Griechische Grammatik 1 (München: Beck).

Stang, Christian S., 1966. Vergleichende Grammatik der Baltischen Sprachen (Oslo: Universitetsfor-
laget).

Sturtevant, Edgar H., 1940. The pronunciation of Greek and Latin. 2nd edition (Philadelphia: Linguistic
Society of America).

Togeby, Knud, 1964. “Review article: Qu’est ce que c'est la dissimilation?", Romance Philology 17:
642—67.

Wanner, Dieter, 1975. “A note on diphthongization", SLS 5 2: 186—202.

Linguistic areas: getting at the grain of history*

Peter E. Hook (University of Michigan)

A perennial problem in the study of linguistic areas is the indeterminate nature of the data
adduced to link observed effect with asserted cause. The present paper reviews four possible
sources of structural parallels in adjacent languages: viz, coincidence, typological harmonics,
genetic relationship, and cultural contact. The kind of data which select cultural contact as
cause (and which exclude the three other possibilities) is identified and instantiated from
field work in South Asia.

Prolonged periods of interaction among speakers of languages belonging to different (or at
least not demonstrably related) families may lead to convergence of structure (or of other
characteristics: see Hook 1982) which is not explainable except by reference to that interac-
tion. This opens up the attractive possibility that we may use the study of linguistic areas
(together with other kinds of tacit evidence) as means to inferring prehistorical cultural
contact among such groups. As an instance of such inference (controlled by independent
historical testimony), the absence of an infinitive in British Romani may be cited:!

kam -av-as te ja-v kher-e (Sampson 1926:IV—131)
want-1S-IMPRF that go-1S home-LOC ‘ want to go home.’
brisindo ja-la te de + (Sampson 1926:11I-191)
rain go-3S that give-3S "It's going to rain.’

This syntactic peculiarity is an indication of protracted residence in the Balkans. We can be
fairly sure of this interpretation since all other forms of Indo-Aryan have an infinitive; the
lack of one distinguishes Balkan languages from their neighbors;? and there are historical
records of the Gypsies’ sojourn there.” In this paper I attempt a preliminary discussion of a
basis for the evaluation of convergence phenomena as indicators of cultural contact when
there is no explicit historical record.

In his landmark study of South Asia as a linguistic area, Colin Masica (1976) makes an
important methodological point: In a linguistic area the parameters along which languages
of different families are similar must in fact define that area. That is, to establish a given
geographical region as a linguistic area we must show that its languages have features in
common which distinguish them as a group from languages outside the area. Thus, for each
feature that he discusses, Masica traces its presence (or absence) in ever widening circles
until he reaches the line that separates languages having it from those that do not. By fol-
lowing this procedure, Masica discovers a very surprising thing, surprising at least to most
Indologists: The syntactic (and semantic) features he selects do not define South Asia as a
linguistic area. Rather they show a commonality in structure that includes language spoken

156

in vast areas of the Soviet Union (the Central Asian Socialist Republics), Korea, and Japan,
an area which Masica has more recently^ termed the ‘Indo-Turanian’ linguistic area.

What are we to make of this? Linguists (at least those linguists who are Indologists) are
familiar with the idea that Indo-Aryan, Dravidian and Mundan languages may have come to
resemble one another as much as they do by virtue of the millennia of close cultural con-
tact for which there is direct historical evidence as well as obvious manifestations in every
sphere of life: religion, philosophy, music, cuisine, etc.” Are the historical incursions of
White Huns and Mughals into India of the same order of magnitude and intensity to ex-
plain a degree of linguistic convergence between South Asian and Central Asian languages
not much different from that found between Indo-Aryan, Dravidian and Munda in South
Asia itself?

In fact, a great deal of attention need not be paid to this question for the same degree of
convergence (or similarity) can be found between the Altaic of the seventh or eighth cen-
turies A.D. (see Appendix A) and inscriptional Tamil of the same or even earlier periods
(Appendix B). Considering the enormous distance separating the site of the Orkhon inscrip-
tions from Tamilnadu and the absence of historical records of contact between Altaic and
Dravidian speakers prior to the sixth century, we must either assume prehistoric contact
(or very remote genetic relationship) to be the explanation or consider whether linguistic
areas may owe their existence to mechanisms other than these. It is the nature of such
‘other mechanisms’ that must be understood before we know how to assess the historio-
graphical importance of the Indo-Turanian linguistic area.

There are two explanations for linguistic areas which do not appeal to contact or to remote
genetic relationship: (1) coincidence, and (2) typological harmonics. The likelihood of
simple coincidence’s being the explanation cannot be determined without a better under-
standing of typological harmonics. That is, unless we know, in general which typological
features pattern (or tend to pattern) with which, we cannot judge what the probabilities
are of a given constellation of features arising purely by chance. For instance, if we assume
that the four features (out of the five used in Masica 19765) that show linguistic parallels
between South and Central Asia are both independent and roughly equipollent (i.e., equally
likely to be present or absent), then the likelihood that a Central Asian linguistic area
causally disconnected from South Asia would show the existing convergence simply by
chance is one in sixteen. If some or all of those four features tend to pattern together in
the world's languages then this rather remote probability of accidental convergence im-
proves. And improves rather markedly. How else are we to explain the presence of lan-
guages of typically Indo-Turanian structure in the western cordilleras of South America
(at the very antipodes of South Asia)? Ecuadoran Quichua, for instance, shares all the syn-
tactic and semantic features with South Asian languages that Orkhon Turkic does (see
Appendix C).

The work of Greenberg and others has shown that certain typological features tend to
cluster together. If a language has a basic word order in which the object precedes the verb
then it is much more likely to have postpositions than it is to have prepositions, auxiliaries
are much more likely to follow their verbs than to precede them, etc. Thus, agreement
among languages along several of these parameters cannot be considered cumulative evi-
dence for establishing a linguistic area, or, to put the matter more precisely, such agree-
ments cannot be considered additive. Rather, a complex formula has to be worked out

157

specifying the weight to be given to each individual parallelism. Recognition of this prin-
ciple is implicit in the organization of Masica’s discussion of South Asia as a linguistic area
where word order phenomena are treated together in a single chapter, as is the “dative sub-
ject’ construction with the absence of a verb ‘to have’. To work out such a formula will re-
quire that we determine ratios of incidence for each feature of interest to us in the total
stock of the world’s languages (or in some controlled sample of them: see Tomlin 1979)
and then, assuming them to be independent, compute the expected frequency of each
combination of them. Comparing these expected frequencies with those actually observed
will tell us how much to discount for typological harmonics, even if there is no satisfactory
explanation for those harmonics in linguistic theories presently available to us.

Let us take a very simple example of this: In Appendix II of his pioneering essay “Some
Universals of Grammar with Particular Reference to the Order of Meaningful Elements”,
Greenberg classifies a largish number of languages according to different combinations of
basic word-order, post- vs. prepositions, adjective before noun vs. noun before adjective,
etc. The listing of languages and groups of languages he surveys totals approximately
thirty-three lines of type. Taking three of these variables individually we find the following
ratios of incidence:

A. SOV word order 13 39% SVO word order 14 42%
B. postpositions 17 52% prepositions 16 48%
C. adjectivenoun 12 36% noun-adjective 21 64%

(Here the integers are the number of lines of type. They do not total 33 for variable A
because I am not considering verb-first word-order here.) Assuming these variables to be
independent of one another, we can calculate expected ratios of incidence for each com-
bination. For simplicity’s sake I will find the values for A into B and for A into C (the
integer following the figure for percentage is the expected number of lines):

I. AB 20% 7, -AB 22% 7, A-B 19% 6, -A-B 20% 7
Il. AC 14% 5, -AC 15% 5, A-C 25% 8, -A-C 27% 9

The observed percentages and numbers of lines for each of these combinations are:

I. AB 36% 12, AR 12% 4, AB 3% 1, -A-B 27% 9
H. AC 18% 6, -AC 12% 4, A-C 21% 7, -A-C 27% 9

Comparison of these two sets of figures shows that the ‘basic word order’ of a language is
much more closely tied to post- vs. preposition than it is to the relative order of adjective
and noun. That is, features A and B tend either both to be present or both to be absent to
a stronger degree than do features A and C. Therefore, agreement on the latter between
languages is more probative in the demonstration of a linguistic area than is post- vs. pre-
position and accordingly should be given a greater weight in the calculation (assuming that
basic word order has already been given a full vote).

However, there is a limitation on this procedure which may be severe enough to dissuade
anyone from undertaking the arduous task of determining separate frequency ratios for a
large sampling of syntactic and semantic traits: It is possible that the world is not big
enough, that the number of coexisting human languages is too small to give a representative
sample of the full potential of man’s language faculty and that cultural contacts are too fre-
quent for us ever to know what the inherent typological characteristics of human language

158

would be if individual languages were allowed to develop undisturbed in pristine isolation.
In short, all languages may belong to a single linguistic area, namely, the Earth.

Such ‘bons mots’ aside, there is another reason for not undertaking mammoth typological
studies of the world’s total stock of languages (interesting as such studies might be in
themselves): From the point of view of demonstrating linguistic areas they are simply not
necessary. Rather if we turn to detailed cross-dialect, fine-grain comparisons we can show a
pattern of convergence that will conclusively put to rest the two-headed bogey of coin-
cidence and typological harmonics and which will also distinguish contact phenomena from
common retentions associated with remote genetic relationships (eg. Nostratic). If a statis-
tical study of a given feature of syntax or semantics shows a regular correlation with dis-
tance, and, especially, if this correlation is independent of genetic groupings and sub-
groupings, then we have a prima-facie case for contact as cause.

Recently I have conducted field research on the geographical distribution of aspectual con-
trasts in South Asian languages. I happened to include a pair of complex sentences on the
questionnaire which show a regular progression in the frequency ratios of subordinate-main
as opposed to main-subordinate clause orders as one proceeds from north to south within
western Indo-Aryan (Sindhi, Kutchi, Gujarati, Marathi). (See figure.) Subordinate-main
order is typical of Dravidian; main-subordinate, of Persian. Since we are no longer dealing
with a simple binary (yes-no, present-absent) value for a feature, but with a finely modu-
lated correlation of frequency with distance, coincidence is ruled out.” Ascribing the facts
to typological harmonics simply puts the operative factor at one or two removes and still
does not explain the regularity in the geographic distribution of that factor. Similarly
remote genetically based groupings and sub-groupings do not vary continuously with dis-
tance: i.e., the Stammbaum model would not predict that Marathi would share more syn-
tactic and semantic features with Kannada and Tamil than the North Dravidian language
Brahui does.

In effect, fine-grain correlations of this kind can be explained only by positing chains of bi-
dialectal and/or bi-lingual speakers along which change in some feature of syntax or se-
mantics is transmitted much in the manner distortions in a message are propagated and ela-
borated in the party game ‘Telephone’. Such correlations promise to do for convergence
studies what the regularity of sound correspondences did for the development of historical-
comparative linguistics. The difficulties in establishing such correlations have to do with
the availability of relevant information? Before Labov linguists were rarely interested in ga-
thering data on the relative frequency of competing forms.

Assuming that such data become available in the future, we may discover that the fine-
grain processes and social interacitons that lead to clines in the frequencies of competing
linguistic forms and structures may be relatively short-lived while their global effects may
persist for centuries after them. For example, the Indo-Turanian linguistic area uncovered
by Masica may be a kind of fossil, a trace of an area once knit together by chains of bi-
dialectal and bilingual speakers, but now subject, in its various parts (and from diverse
directions), to new influences that have yet to obscure the large-scale parallels in structure
established in prehistoric times. An ambitious survey of competing patterns in a number of
linguistic dimensions carried out on a village-to-village scale along selected geographic arcs
in South Asia and Central Asia is needed to answer this question.

159

i Kandahar

® Quetta
OO
O

è Kalat,

Sukkur g
O d
O o(o

o NO UN

pend y [^
Karachi a S
Y à

Bhu 3 Ahmadabad /®.

e Jodhpur

e Turbat

Oo c

Porb Is
Percentage of returns SE

showing subordinate
before main clauses:

Balochi, Pashto, Brahui...00%
Sindhi........ eee eere ell
Western Gujarati, Kutchi..44
Southern Gujarati.........67
Marathi, Konkani.......... 75

Kannada, Tamil, Telugu
and Malayalam.......100

Key:
= subordinate before main in translations of
‘Wait here until he gives you the letter.'

subordinate before main in translations of
‘I was afraid that you might give him the letter.

= subordinate before main in translations of both.

oe ® ©

= main before subordinate,

Chart of distribution of survey questionnaires classified
according to the relative order of main and subordinate clause.

160

List of Abbreviations

ACC re ae aus accusative
MOSE ee N a adjective
CAUS 22 Sue nahe causative
CIS 2 ni ben cislocative
CP ee ela res conjunctive participle
DAT: zu. 0.4.0 DE i dative
EMPH 2 2.05 we area emphatic
EE ee ee ergative
GENE a are a ee genitive
HORT 2... 22 nn a hortative
IMPER... 22. 2a 42.535 imperative
IMPRE i. Se 250, Soin dew eee, na imperfect
INF" oriana he lea infinitive
INSTRU................ instrumental
LOC-+3f hier ilaele Les ia locative
Nelle none ea NIA noun
Notes

NE asiago noun phrase
OB See inc toa re Wie ea oblique
Pi. ee ee + + + + plural
PB arte denen possessor
PRES. «uu ee .u ei ates present
PST. uuu edic oss past
RSR pub sm ii EH past participle
Reli 155a Temi relative phrase
S te ad ute pete ao Ie Ceci eet NOIRS E E" en singular
STAT EE singular
SM «deer a RR NS standard-marker
SOM ode GLEN EGLA subject-object-verb
Ti ene dui ERE eS mox e eR gu ie tonral
head de" Dean Sonde first person
Eeer e SIN S second person
3 Xue quw UR RUM epus third person

* This paper derives from one presented at the thirty-fifth annual meeting of the Association for Asian
Studies (San Francisco, 1983) as part of a panel on South Asia as a linguistic area. The field work on
which it is based was carried out in India and Pakistan starting in 1978 on research trips supported in
varying degrees by the Horace H. Rackham School of Graduate Studies, University of Michigan; the
American Institute of Indian Studies; and the United States Educational! Foundations of India and
Pakistan (Fulbright-Hays Faculty Research Abroad). Their financial and organizational assistance is
gratefully acknowledged as is the help and co-operation of colleagues at the University of Delhi, the
Central Hindi Institute (Agra), the Northern Regional Language Centre (Patiala), the University of
South Gujarat (Surat), the Oriental College at the Unversity of the Panjab (Lahore), the Institute of
Sindhology (Jamshoro), the Pakistan Information Department and many other institutions and
individuals in both countries. I am particularly indebted to Mrs. Kathleen Chaudhry of Lahore for
her constant help, introductions and hospitality and to Dr. H.K. Gaur of Surat for providing much

needed data from Western Gujarat.

The first South Asians to reach the shores of England arrived well before the East India Company

was founded, at the latest by 1500 according to Vesey-Fitzgerald (1944:20). Their northwesterly
migration began sometime before 250 B.C. (Turner 1927:23)

2
Modern Greek: Oeo va ma-w
I-want that go-1Sg
Bulgarian: toj ne iska da dojd-e
he not want that come-3Sg
Rumanian: vreau sa cint
I-want that I-sing
Albanian: dua të shkruaj
I-want that I-write
Macedonian: saka da jarexava
I-want that it I-solve
Grecanica-Bova: /6el-a na mu fer -i
Calabrese: vuli-a mu mi pòrt -a
n he-wanted that me bring-3SG

(1944:6ff.).

See Sandfeld (1930) and Solta (1980). The Balkan languages that lack the infinitive include:

‘I want to go.’

‘Il ne veut pas venir.’
(Beaulieux 1950:326)
‘Je veux chanter.’
(Cazacu 1967:125)

‘I want to write."
(Solta 1980:212)

‘He wants to solve it.’
(Lunt 1952:85)

‘He'd like to bring me.’
(Solta 1980:214)

For an Arab report of Gypsies among the Byzantines in the ninth century A.D. see Vesey-Fitzgerald

161

In his paper entitled “South Asian languages: typological coincidence or areal convergence?” present-
ed at the 35th annual meeting of the Association for Asian Studies, 1983.

For a first look at such convergence in folk narratives, see Hook (1979).

The fifth one Masica discusses, the ‘dative subject’, does not exist in Central Asian languages.

The probability of coincidence’s being the ‘cause’ in the appearance of a regular progression in a
series of geographically ordered data sets is equal to £, where n is the number of sets. In this particu-
lar instance, where there are six such sets, the chances of their exhibiting such a progression purely
by coincidence are twoin6x5x4x3x2x 1 or one in three hundred sixty (0.0028).

Relevant information has sometimes been collected, by accident as it were. For an attempt to ex-
tract such information from Grierson’s linguistic survey of India, see Hook (1977).

Part of Deshpande's argument for a Dravidian source for retroflexion in the Rgveda appeals to geo-
graphical information (1979:252ff). Southworth 1974 goes considerably further in his use of data
from Grierson to infer, on the basis of the incidence of retroflexion in modern Indo-Aryan, that
Maharashtra, Gujarat and Sind are the areas most likely to have been Dravidian-speaking at the time
of the Aryan incursions into South Asia. The problem with this is that the present distribution of lin-
guistic features may have nothing to do with the situation that obtained three to four thousand years
ago. For an analogous difficulty in attributing the gorgia toscana to an Etruscan substratum see
Kurath (1972:154—156).

162
Appendix A: Orkhon Turkic

Orkhon Turkic, as known to us from inscriptions first discovered 250 years ago in what is
today central Mongolia, shares with Dravidian the four syntactic and semantic parallelisms
that Masica 1976 uses to define the Indo-Turanian linguistic area: (1) word order, (2) de-
rived causative verbs, (3) conjunctive participles and (4) explicatior compound verbs.
Examples of these (unless otherwise indicated, from the eastern or southern faces of the
Kiil Tigin inscription) are:

(1) word order:
a tutuq -uy dlig -in tut-di (E38) SOV
governor-ACC hand-INSTRU captured
‘(He) captured the governor by (his own) hand.”
b türük qara qamay bodun (E 8) AdjN
Turk black entire people
‘... the Turkish common people. . .°

c ini -m kül tigin birlà (E 26) Postposition
brother-my Kul prince with
‘... with my brother, Prince Kul...

d Otükän yis -da yig idi yoqürmis (S4) SMAdj

Otukan mtn-from better at-all not is
‘There is no better (place) than the O. Mountains.’
e umay tig ög -üm (E31) RelP-NP
Umay like mother-my
‘My mother (who is) like the Goddess Umay. . .°

(2) derived causatives (and transitives):
a öl- ‘die’ ölür- ‘kill’; kal- ‘come’ kaliir- ‘bring’
b dr- ‘be’ artiir- ‘make’; yarat- ‘make’ yaratur- ‘have made’
c anar adinč(č)iy barą yaratur -t -um

them (DAT) wonderful tomb build-CAUS-PST-1Sg
‘I had them build an extraordinary mausoleum.’

(3) The affix of the conjunctive participle (CP) in Orkhon Turkic has a number of forms.
Chief among these are a vowel (-a/-d, -i/-i, -u/-ü) and an ending -p or -pan/-pän. Like
their counterparts in Dravidian (and unlike English having V-ed) the Orkhon conjunc-
tive participles show great syntactic freedom occurring in imperatives and questions (as
well as in the narrative declaratives where analogous constructions occur in English and
most languages):

a yaragli gantin käl -ip yań -a eltdi (E 23)
armed whence come-CP disperse-CP sent
‘From where did the armed (ones) come and scatter (you)?
b buni körü bil -in (S12)
this (ACC) see-CP know-IMPER
‘See this (inscription) and learn (from it)!’
The same-agent constraint (Hook 1976) applies (although such agents do not have to
be the same in surface structure):

(4)

(5)

163

c yogla) aruat yät-à yadayin iyal tutunu ay -turt -um
upwards horses lead-CP on-foot trees cling -CP climb-made-I
‘I had them climb upwards on foot leading their horses and clinging to the trees.’
(Tonyukuk I, N 1)
(Here, it is clear from context that the leader of horses and the clinger to trees is not
the T of the account.)

Explicator compound verbs in Orkhon Turkic, as in Dravidian, are formally indistin-
guishable from conjunctive participle plus finite verb constructions (cf. 3a):
a uluy irkin azginaär -in täz -ip bard + (E 34)

great (name) few men-INSTRU flee-CP go -PST-3Sg

‘The great Irkin ran away with a few (of his) men.’

The explicators are homophonous with lexical verbs:

b bodun yer -in sub -in id -ip tabyac-yaru bardi
people land-their water-their leave-CP China -toward went
‘... people left their land and water and went toward Ching.
(Bilgä Kagan incription, E 35)

c qayanin yitürü id -mis (E 7)
kagan-their lose -CP leave-PST
‘.. . (they) lost their kagan. . .°

The alternation of the explicator with its absence (one of the two criteria for the clas-
sification ‘compound verb’; see Hook 1977:336) is found in Orkhon Turkic under
similar conditions:
d kül tigin qon vil -qa viti vegirmi-ki uc-di (KT NE)

Kul Prince Sheep year-LOC seven twenty -LOC fly-PST

"Prince Kui died in the Year of the Sheep on the 17th day.’

Here the tone is matter-of-fact. Contrast this with:
e uĉa bord -iy-iz (KT SE)

fly-CP go -PST-2 -P1
where the writer, Prince Kul's nephew, expresses his sense of loss (The sentence follows
directly on: ‘You used to nourish (the people) better than your beloved children and
descendants.’).

Another feature common to Orkhon Turkic and Dravidian of the time is the use of
participle of the verb ‘say’ (te-/ti-) as a complementizer with verbs of speaking and
thinking and to introduce clauses of purpose (cf. Kuiper 1967):

a oluruy ti -yin te -mis ‘“Stay!” he said.’
stay-2PL say-ing say-PST (Tonyukuk I N 10)
b üküs te -yinnä -kä qorqur biz (Tonyukuk H W 4)

many say-ing what-DAT fear -PRS we
‘Why should we be afraid (of there being) many (of them)?”
c arqi id -mazti -yin sülä d -im (Bilgà Kagan E 25)
caravan send-not say-ing campaign-PST-1Sq
‘I went to war (against them) because they did not send the (tribute) caravan.’

164
Appendix B: Inscriptional Tamil

The earliest Tamil prose of appreciable length is that found some fifty years ago in an
inscription of some forty lines that was unearthed in Pallankövil, a small village in Tanjore.
The text, recording a grant of land, is considered to date from before 550 A.D. (Zvelebil
1964:6). From it we may illustrate the same syntactic and semantic features as have been
exemplified in Appendix A.

(1) word order:

a vajranandi-kkuravar-kk -e... nilan kotu-tt -om (54—56) OV
(a name) teacher -DAT-EMPH land give -PST-we
‘To the same teacher V. ...we gave land...’

b tan nat... (29) AdjN
your district...

c peru -nänk-ellai-yaka -ttu (47-48) Postposition

great-four -limit-T-interior-OBL
‘... within the great four boundaries. . .'
d (no comparatives available in the inscription)

e kanru-mey pal (49) RelP-NP
calf -graze portion
‘. .. the part (where) the calves graze. . >?

(2) derived causatives:

While no clear instances are available from the Pallan-kovil inscription, they may be
found in earlier inscriptions (from 200 B.C.) found in caves:
a siri vakarucanataritan kotu-pi -t -on (Arittapatti II, Zvelebil 1964:552)
shri (name) (name) cut -CAUS-PST-3Sg
‘Shri Yakaru Canataritan . . . had (this cavern) cut.’
Compare this with the non-derived transitive in:
b velatai nikamator kotior (Arittäpatti IL, in Zvelebil 1964b:552)
(name) citizens (?) cut
‘The people of Velatai cut (the cavern).'

(3) conjunctive participles:

a nattar -um tirumukan kan-tu tolu-tu... (33—34)
members of natu too royal-order see -CP namaskrtya
‘And the members of the district assembly having seen the royal proclamation and
having shown their deference. . ?

(4) As in Orkhon Turkic the morphology of the explicator compound verb construction is
indistinguishable from that of conjunctive participle plus finite verb:
a arai -y-olai cey -tu kotu-ttu vitu -taka... (32-33)
speak-T-leaf make-CP give -CP leave-HORT
‘... prepare the proclamation and issue (it).'
Note the use here as in Orkhon Turkic of a verb homophonous with the verb for ‘leave’
as explicator.

165

(5) The conjunctive participle of the verb en ‘say’ is used as a sentential complementizer
(see Kuiper 1967):

a tahkal um...araivolai cey -tu kotu-ttu vitu -taka -v-en -ru
you-PL-too speak-leaf make-CP give -CP leave-HORT-T-say-CP

nattar -kku t-tirumukam vit -a
members-DAT T-royal-order leave-INF

(31-33)

‘(The king) has given an order to the members of the assembly that they too should

... prepare a proclamation and issue (it).

Appendix C: Marathi and Quechua

Marathi is one of the Indo-Aryan languages most strongly Dravidianized in its syntactic and
semantic structures. The examples adduced here show structural parallels between it and
Quechua, the modern descendant of the state language of the pre-Columbian empire of the
Incas. Convergence through cultural contact may in this instance be safely ruled out.

(1) word order:

(2)

mi tya-là paise -@ de + -o
ñuka pay-man kullki -ta ku 4 -ni
I him-to money-ACC give-PRES-1Sg
Tl give the money to him.’

sundar stri

sumak warmi

pretty woman

baban- “<a mitrà -či Seti -¢
tayta a -pa amigum pa Cakran
father-POSS-GEN friend -POSS-GEN farm -POSS
'(my) father's friend's farm. . .'

majha-hun — huar

noca -manta amauta

me -from smart

‘wiser than me...’

huanni di -lela patr -¢ pathav
juan cu -shca quillca-ta cachai
Juan -ERG give-PSTP letter -ACC send
‘Send the letter that Juan gave.”

SOV
(Muysken 1977:19)
AdjN

Postposition
(Parker 1976:91)

SMAdj
(Markham 1972:28)

RelP-NP

Sentential objects, in both Marathi and Quechua, often follow their verbs (see sentence
5-a below).

derived causatives (and transitives):

mar- mår- mdr -av-
wañu- wañu-ti- wañu-Ci-ti-
die kill kill -CAUS

(‘have someone kill”)

166

(3)

(4)

(5)

(6)

(7)

conjunctive participles:

a gharà -lā pohotsun hosejeva -lā
wasi- -man chaya -shpa huzi miku-rka (Muysken 1977:29)
home-OBL-to arrive -CP Jose eat -PST
*( After) coming home Jose had his meal.’

As in Altaic and in Dravidian, the explicator compound verb construction is based on
the conjunctive participle form of the main verb in both Marathi and Quichua (Quichua
is the Ecuadoran form of Quechua). Explicator constructions according to Bruce Mann-
heim (personal communication) are not to be found in Peruvian and Bolivian dialects
of Quechua.
a vatsun de

cati -shpa cuy (Vazquez 1940:126)

read-CP give

‘Read (this) (for me).’

In the Sierra this construction has entered Spanish as a polite imperative or request:

b da -mele -y-endo esta carta (Toscano Mateus 1953:284)
give-me read-T-ing this letter
Please read this letter for me.’
In Quichua the construction, apparently limited to transitive verbs, may include ex-
plicator auxiliaries homophonous with the verbs for ‘give’, ‘leave’, ‘put’ and ‘throw’.
The system appears more extensive in Marathi (see chart in Masica 1976:146).

Both Marathi and Quechua use the conjunctive participle of a verb ‘to say’ as com-
plementizer:
a ma-lä kala-vi dd “le sukh-rup pohots-ó -lo mhan-un
ni chi -mu-wa-rga -nallinta chaya -mu-ni ni -shpa

me-to tell -CAUS-CIS-me-PST-3 safely arrive -CIS-1Sg say -CP

‘He informed me that he had arrived safely.”
b bagh-at ahe zhad-an at pakfi-@ pakad-ay-tsay mhan-un

riku -u -ni yura-una-ma pisku-ra api -nga-k ni -sha

look-ing-1Sg tree -PL -LOC bird -ACC grabbing for) say -CP

‘I’m looking in the trees in order to catch a bird.’

Marathi and Quechua both have a 'dative-subject' construction (although as Masica
1983 has noted, this is substantially less developed in Quechua):
a ma -la zhop-aytsa ah-e
nuka-ta punu-naya -ó -n (Muysken 1977:123)
me -to sleep-want be-3Sg
7] feel like sleeping.’

This is a feature South Asian languages do not share with those of Central Asia.

Emeneau (1974) shows that Sanskrit api and Dravidian Am have the same semantic
range: ‘also’; ‘and’ (in a series); ‘even’, ‘although’; as a totalizer (‘all six’, ‘both’, etc.);
and as an indefinitizer: kopi ‘whoever’ = kah ‘who’ + api ‘also’. Most of this can be
shown for Quechua pish:

167

a gel -o hi
ri -rca -ni pish
go-PST-1Sg also
‘(and I) went, too.’

b kay-mam-pis wak -mam-pis (I cannot find this pattern in Marathi.)
this way also that way also
‘...this way and that...’ (Parker 1976:146)
c pïña-ri -c -pi -pish mana ri -sha-chu rag ye -unhi kay upyog
anger-go-er-on-also not go-will-no anger come-CP also what use
‘Even if he gets angry, I won't go.’ ‘Even if you get angry, it’s no use!’

d Quechua: pi ‘who’, pi-pis ‘somebody’; ima ‘what’, ima-pis ‘something’
Marathi: kon ‘who’, kon-i (kon + hi) ‘somebody’, koni hi ‘anyone’
kutha ‘where’, kutha hi ‘anywhere’
kay ‘what’, kahi ‘some’, ‘something’

e pats-hi gele (absent in Quechua?)
five -also went
‘All five went.’

References

Beaulieux, Léon, 1950. Grammaire de la langue bulgaire (avec le concours de Stefan Mladenov) 2 édi-
tion, révisée et corrigée (Paris: Institut d'études slaves).

Cazacu et al. (ed.), 1967. Cours de langue roumaine (Bucarest: Editura Didactica si Pedagogica).

Deshpande, Madhav M., 1979. “Genesis of Rgvedic retroflexion: a historical and sociolinguistic investi-
gation”, in: Aryan and non-Aryan in India, ed. by Deshpande and Hook (Michigan: Ann Arbor),
235-315.

Deshpande, Madhav M./Peter E. Hook (eds.), 1979. Aryan and non-Aryan in India (Center for South
and Southeast Asian Studies, University of Michigan: Ann Arbor).

Emeneau, M.B., 1956. “India as a linguistic area”, Language 32 1:3-—16.

Emeneau, M.B., 1974. “The Indian linguistic area revisited”, in: Southworth and Apte (eds.) Contact
and Convergence in South Asian Languages IJDL vol. 3 No. 1 92-134.

Greenberg, Joseph H., 1956. “Some universals of grammar with particular reference to the order of
meaningful elements”, in: Universals of Language, edited by Joseph H. Greenberg (Cambridge: M.IT.
Press), 73--113.

Hook, Peter E., 1976. “Astädhyäyi 3.4.21 and the role of semantics in Paninian linguistics”, in: Papers
from the Twelfth Regional Meeting (Chicago: Chicago Linguistic Society), 302--3 12.

Hook, Peter I:., 1977. “The distribution of the compound verb in the languages of North India and the
question of its origin”, in: International Journal of Dravidian Linguistics 6:336--351.

Hook, Peter E., 1979. “The marriage of heroines and the definition of a literary area in South and
Central Asia”, in: Deshpande and Hook (eds.) Aryan and non-Aryan in India (Michigan: Ann Arbor),
35-54.

Hook, Peter E., 1982. “South Asia as a semantic area: forms, meanings and their connections”, in:
South Asian Review ed. by P.J. Mistry, vol. 6, No. 3, 30-41.

Kuiper, F.B.J., 1967. “The Genesis of a Linguistic Area”, in: Indo-Iranian Journal, 10, 2/3: 81-102
(Reprinted in Southworth and Apte (eds.) 1974: 135-153).

Kurath, Hans, 1972. Studies in Area Linguistics (Bloomington: Indiana University Press).

Lunt, Horace Gray, 1952. A grammar of the Macedonian literary language (Skopje: Drzh. kn-vo).

Markham, Clements R., 1972. Contributions toward a grammar and dictionary of Quechua, the language
of the Incas of Peru (reprint of 1864 edition) (Osnabriick: Biblio Verlag).

Masica, Colin P., 1976. Defining a linguistic area: South Asia (Chicago: University of Chicago Press).

168

Masica, Colin P., 1983. South Asian languages: typological coincidence or areal convergence? Presented
at the 35th annual meeting of the Association for Asian Studies.

Meenakshisundaram, T.P., 1965. A history of Tamil language (Poona: Linguistic Society of India).

Muysken, Pieter, 1977. Syntactic developments in the verb phrase of Ecuadoran Quechua (Lisse: Peter
de Ridder Press).

Parker, Gary J., 1976. Gramdtica quechua: Ancash-Huailas (Lima: Ministerio de Educación).

Sampson, John, 1968. The dialect of the Gypsies of Wales, being the older form of British Romani
preserved in the speech of the clan of Abram Wood (Oxford: Clarendon Press).

Sandfeld, Kristian, 1930. “Linguistique balkanique; problèmes et résultats”, in: Collection linguistique
31. (Paris: Société de linguistique de Paris).

Solta, Georg Renatus, 1980. Einführung in die Balkanlinguistik (mit besonderer Berücksichtigung des
Substrats und des Balkanlateinischen) (Darmstadt: Wissenschaftliche Buchgesellschaft).

Southworth, F.C., 1974. "Linguistic Stratigraphy of North India", in: Contact and Convergence in
South Asian languages, edited by Southworth and Apte, 201—223.

Southworth, F.C. and M.L. Apte (eds.), 1974. Contact and Convergence in South Asian languages. A
special issue of International Journal of Dravidian Linguistics vol. 3, no. 1.

Tekin, Talat, 1968. A grammar of Orkhon Turkic. Uralic and Altaic Series vol. 69 (Bloomington: In-
diana University Publications).

Tomlin, Russell, 1979. An Explanation of the Distribution of Basic Constituent Orders, University of
Michigan dissertation.

Toscano Mateus, H., 1953. El español en el Ecuador: Revista de philologia española, anejo no. 61
(Madrid).

Turner, Ralph Lilley, 1927. “The Position of Romani in Indo-Aryan”, Gypsy Lore Society Monograph
No. 4 (London: B. Quaritch).

Vazquez, Honorato, 1940. Reparos sobre neustro lenguaje usual (Quito: Editorial Ecuatoriana).

Vesey-Fitzgerald, Brian S., 1944. Gypsies of Britain: an introduction to their history (London: Chap-
man and Hall Ltd.).

Zvelebil, Kamil, 1964a. Tamil in 550 A.D.; an interpretation of early inscriptional Tamil (Prague:
Publishing House of the Czechoslovak Academy of Sciences).

Zvelebil, Kamil, 1964b. “The Brahmi hybrid Tamil inscriptions”, in: Archiv Orientální 32:547—575.

Mantra glosses in the Satapatha Bràhmana: more light on the development of
the Vedic verbal system

Stephanie W. Jamison (Cambridge, Massachusetts)

The Satapatha Bràhmana is in great part a vast commentary on the mantras of the Väjasa-
neyi Samhita (= White Yajurveda). As such, it often quotes a mantra directly and then
glosses or paraphrases it, wholly or in part, sometimes with a gloss pure and simple, as in
1.3.4.16:'

dhruva asadann iti dhruva hy äsadan
‘Firm they have sat down’, for firm they have sat down.

more often by describing the actions appropriate to the speaker of the mantra or the moti-
vation for its use, as in 1.1.3.11:

sd próksati agndye tva jüstam prôksämifi
He sprinkles (saying) ‘I sprinkle thee, enjoyable to Agni’.

As these examples show, the prose paraphrase is generally as close as possible to the wording
of the mantra, with technical adjustments in person and the like, necessitated by the switch
from the putative speaker of the mantra to the commentator on it. Novelty and variety of
phrasing are not valued for themselves.

Given this faithfulness, it is in fact surprising how often the prose expression deviates from
its text. These deviations are all the more interesting and valuable because they must result
from pressures strong enough to overcome the counterpressure to repeat the mantra exact-
ly. These discrepancies provide an almost ideal window on the development of the Vedic
verbal system, and one that has been surprisingly neglected, as Renou pointed out almost
30 years ago (Histoire de la langue sanskrite (1956) 47, n.2). I know of only one brief
treatment — that of Jayasuriya in the Ceylon University Review (1957). Though useful,
it is hardly systematic, and, moreover, not easily accessible to many Western scholars.

This paper will examine the deviation in verbal forms, using a complete collection of the
glossed verbs in the SB (Madhyandina recension). At first glance, the data seem bewildering,
almost intractable: the problem is that glossing practice is remarkably inconsistent. Obso-
lete forms are sometimes glossed by their modern replacements, sometimes left unglossed.
So, pati ‘protects’ is glossed by gopayati ‘id.’ 15 to 20 times (e.g. 1.5.1.22, I.9.1.20), but
appears in unglossed mantras about an equal number of times (e.g. 1.8.3.19, 1.9.2.17).
Forms which remain at least marginally alive in prose are sometimes replaced when they
appear in mantras, sometimes not, sometimes glossed by themselves. For example, bharati
is glossed by harati 4 times (e.g. 111.2.2.25, V1.6.3.8), but is commonly glossed by itself

170

(e.g. VI.4.4.8) or left unglossed (e.g. 11.4.2.15). Bharati is rarely used independently in
prose, but does occasionally appear (e.g. 1.4.2.2, where it was induced by vocative bhärata;
VIII. 1.1.9, induced by bAharádvaája-). Indeed, of forms of the same morphological class,
parallel to each other in the same passage, one may be modernized, the other not, as with
the s-aorists in IX.2.3.20:

yäksad agnir devö devamı a ca vaksad iti yäksac caivägnir devö devan ä ca vahatv ity
etat

‘Agni the god shall worship the gods and convey (them) hither’, that is ‘Agni the god
shall both worship the gods and let him convey (them) hither’.

The picture seems to be one of a verbal system in flux: where even obsolescent forms
remain within the passive command of the speech community and can often still be used
actively, though perhaps with self-conscious archaism.

However, despite these inconsistencies, certain patterns do emerge — patterns of two types.
On the one hand, it is possible to assess the relative stability of categories within the verbal
system, and of individual roots and stems. On the other, certain kinds of glossing practice
cluster in certain portions of the text: in particular a large number of these practices are
confined to, or far more common in,the Sändilya kändas VI—X, and absent, or almost
absent, from the Yajfiavalkya kandas. I will examine these two types of patterns in turn.

With regard to verbal categories, the general picture is consistent with the common opinion
on the development of the Vedic verbal system; surprises may lie in the relative strengths
of the various categories. Two categories of relative stability are voice and inflectional
endings. Substitutions of voice are very rare. Of those that do occur, there are a very few
apparently unmotivated substitutions, as in VII. 2.2.5:

yunákta sirà ví yuga tanudhvam (ti yufijánti hí siram ví yugani tanvánti?

But most changes in voice are an automatic result of substituting a stem with one inherent
voice for one with its opposite; for example, in VI. 7.2.2 the active aorist vy ddyaut is
replaced by medial present vidydtate, since the aorist is only active, the present only
medial. À special case of this is the occasional replacement of the old athematic intensive
with the newer -ydte type, as in VI. 4.4.7, with part. kanikradat glossed by kanikradya-
mänah.?

In inflectional endings, there is some tidying up of earlier alternative endings, e.g.

VII. 5.2.39 srnudhî > srnü
VI. 8.1.6 juhotana > juhuta
VIL. 3.1.22 bharamasi + haramas -

but these changes are rare. Irregular nominal endings seem more likely to be replaced (e.g.
neut. pl. -4 by -ani).

More surprising is how rarely one alternative stem is substituted for another within one
aspect — i.e. one type of present for another, one aorist for another — despite the plethora
of multiform presents and aorists in Vedic. This happens occasionally in the aorist with
short, ambiguous second and third person forms, which are converted into longer forms
with clear personal endings:

171

1.2.2.15 (= 11.94.18) bheh > bhaisih

VI.3.3.14 dkar > dkarah
VII. 1.1.43 abhar — abharsit
XIV. 2.2.16 ayat > aydksit

In the present it hardly occurs except (1) in cases of phonological change — loss of occlu-
sion in voiced aspirates (bAárati ^ hárati discussed above; grbhnati > grhnäti 4times, e.g.
VI. 4.4.17); (2) in cases involving -dva- formations, usually when the -dya-formation ap-
pears to constitute a new root (eg. srindtu > srapáyati TII. 8.3.20; prnati > purayati e.g.
VI.5.4.12). The major exception here is the root kr, whose old stem Arnoti, krnuté is
completely obsolete in the SB; karóti, kuruté has taken over its functions. In the glosses as
well, karóti can replace krnoti (e g. 1.9.2.20 krnu > kuru). Yet in fact krnóti often remains
unglossed in the mantras in which in appears: karóti glosses it only 7 times.

Far more common is the substitution of one aspect for another. Often this is semantically
or situationally motivated: the description of the ritual requires an expression of present
time, and the mantra contains a perfect or an aorist, e.g.

IV. 6.9.12 ávidàma deván iti vindanti hi devan

"We have found the gods', for they find the gods.

III. 6.4.14 ayám hí tvà svádhitis tétijànah pranináya ... íty esá hy énam svádhitis
téjamanah prá nayati

*For this very sharp ax led thee forward' (he says), for this sharp ax leads him for-
ward.

However, even when these pragmatically conditioned forms are left aside, substitution of
aspect remains common. Most of these changes involve modal forms and participles, and
the prose substitute is almost always a present. The trend in the language is to restrict non-
indicative forms to the present stem; these data show how far this restriction has advanced
in the SB, as Jayasuriya already pointed out. Compare the following examples, where no
inherent semantic difference between the aspects can be found:

Imperative:

134.12 átvà vdsavo rudrá aditydh sadantv fti . . . eté tvásidantv ity evaitàd aha
‘Let the Vasus, Rudras, and Ädityas sit on thee’, i.e. he says by this ‘Let them
sit on thee’.

IV.1.1.22 mänas tvästv iti prajdpatir vat manah prajapatis tvasnutam ity evaitàd aha
‘Let mind reach thee.” Mind is Prajapati. Thus he says by this ‘Let Prajäpati
reach thee’.

Subjunctive:
IX.2.3.42 yathäa deva ihagämann (ti ydtha deva ihagdchan ity etdd

‘So that the gods shall come hither’, that is ‘so that the gods shall come hither’.
Participle:

V1.7.2.2. drsänö rukma urvya vy ädyaud iti drsyämano hy esá rukmá urvyd vidyótate
‘Appearing as an ornament, widely he has flashed forth', for appearing as an
ornament, widely he flashes forth.

172

However, the case is strikingly different in indicative forms where no present-time expres-
sion is needed. All three past tenses (imperfect, aorist, and perfect) are relatively stable,
and substitutions, when they occur, are chaotic, without a clear direction. Though aorists
and perfects may be replaced by imperfects, the change also goes in the other direction,
and aorists and perfects also interchange with each other. Following are examples of all of
these possibilities:

Aorist > imperfect

VII. 5.2.36 ajó hy ägner äjanista sókad iti yád vat prajapateh sokad äjäyata tád agnéh
‘For the goat was born from Agni's heat.' Truly, what was born from Prajäpati’s
heat, that (was born) from Agni's.

Perfect ^ imperfect

VI.7.4.3 divds pári prathamám jajfie 'gnir iti prànó vai divdh pranad u va esd prathamam
ajayata
*Agni was born first from heaven.’ Heaven is really breath, and he was indeed
born first from breath.

Imperfect perfect

V.1.4.8 gandharváh saptávimsatis té "gré 'svam ayufijann iti gandharva ha va agre 'svam
yuyujur
"The 27 Gandharvas yoked the horse at first.' Indeed the Gandharvas yoked the
horse at first.

Imperfect — aorist

IV.4.4.11 yam avahah usato deva devan... iti... yan devan äväksih
‘Which eager gods, o god, you ‘conveyed hither ... ie. ‘which gods you have
conveyed hither...’

Aorist > perfect

VII.5.19 int samudrant sim asrpat svargan iti... tan esd... anusdth sasarpa
‘He crept over the 3 heavenly oceans’. . . he crept over them.

Perfect + aorist

IV.4.4.10 suga vo devah sadana akarma ya äjagmedam sdvanam jusänä iti sugani vo deväh
sädanany akarma yà ägantedam sdvanam jusand ity evattád àha
‘We have made pleasant seats for you, o gods, who came to this pressing, en-
joying.’ He says by this ‘... who have come...’

As far as I can see, no narrowed semantic spheres have been established for the three dif-
ferent tenses; they remain in competition throughout their ranges. Even the aorist retains a
large domain: recall the examples cited above of ambiguous aorist forms glossed by other
aorists, not only injunctives in ma prohibitives, but also augmented indicatives. And aug-
mented aorists often gloss themselves. For examples dsadan in 1.3.4.16 quoted at the
beginning of the paper, asadat (VI. 4.4.17), asadas (VI. 8.2.6); in contrast we just noted the
aorist imperative dsadantu (I.3.4.12), close to the aforementioned dsadan, replaced by a
present imperative.

173

In mood, clear directions can again be perceived. Though not particularly common, sub-
stitutions in mood are almost always explicable, serving to set up or reinforce semantic
distinctions between the moods, which had earlier been more fluid. (Jayasuriya comes to
the opposite conclusion, but he seems not to have examined the passages carefully.) The
major trend is to replace the optative or the subjunctive with the imperative (or, in the first
person, the old subjunctive) in expressing a wish or desire (sometimes specifically charac-
terized as an asis- ‘hope’):

1.9.1.12 asyam rdhed dhötrayam devamgamayam iti / asyam radhnotu hötrayam
devamgamayam ity evaitad dhasäste 'ydm ydjamanah
‘Might he succeed in this oblation going to the gods.’ He says by this “Let him
succeed . . .". Thus does the Yajamàna express his hope.

1.3.4.2. sám ahám ayvusà sim vdrcasà sim prajdya sém raydsposena gmisiyéti . . .ahám
etaih sam gacha ity evaitád aha
‘Might I be united with long life, with splendour, with offspring, with thriving
of wealth.' By this he says ‘Let me be united with these”.

These mood substitutions often interact with the tendency to limit modals to the present
stem — hence, for example, present imperative/subjunctive sám gachai for aorist optative
sám ... gmisiya in II. 3.4.24; present imperative vahatu for aorist subjunctive vaksat in IX.
2.3.20 (quoted p. 170 above); present imperative radhnotu for aorist optative rdhet in
1. 9.1.12.

While the imperative becomes the mood appropriate for wishes, the optative becomes
especially characteristic of expressions of obligation and necessity. So, in 1. 5.1.26 first-
person subjunctive avahani, now felt as an imperative, is replaced by optative ydheyam in
such a context: |

visve devah sastina ma ... pra me bruta ... vdtha vo véna patha havyám à vo
váhániti .. . ánu má sasta yatha vah . . . havyám váheyam iti

‘O Visve Devas, instruct me, proclaim to me how (and) by what path I shall bring
you your oblation’, (that is) ‘instruct me how I should bring you the oblation'.

Complete change in syntactic construction is relatively rare. The clearest examples involve
the elimination of a dative infinitive or quasi-infinitive by a recasting of the sentence:

WH. 2.2.22 prabüdhe nah punas krdhiti yatha ... prabudhyamaha evdm nah kurv ity
evaitad aha
‘Make us awake again.’ He says by this ‘Make (it) so for us that we shall awake’.
IX.5.1.48 imám yajfiám no naya svar devesu gantava iti... svargam lokám gamayéty etád
‘Lead this worship of ours to go to the sun, among the gods.’ That is, ‘make (it)
go to the heavenly world’.

Note in HII. 2.2.22 also the replacement of the aorist imperative krdhi with present impera-
tive Kuru, in conformity with the usual trend.

The most common type of substitution encountered in the SB is that of one root for an-
other, but this practice can be further subdivided. There are isolated substitutions, scattered
throughout the text. Some of these seem to result from misunderstanding of the glossed
form, especially if it is of very low frequency in the Brahmanas, as in I. 4.1.27:

174

ichà deva vivàsasiti . . . etán no gamayéty evaitád dha
‘Thou dost seek to win (this), o god.’ ‘Make us go to it he says by this.

However, such replacements could simply result from a freer paraphrase than we generally
meet with. Other isolated substitutions are found where no possibility of misunderstanding
exists, e.g. IH. 9.3.5:

urdhväm imäm adhvarim divi devésu hotra yachéti ... urdhvam imam yajñäm divi
devésu dhehity evaitad aha

‘Hold erect this ceremony, the oblations in heaven among the gods.’ He says by this
‘place erect this worship in heaven among the gods’.

(Note here the unusual substitution of an aorist imperative for a present.)

In addition to isolated substitutions, there are systematic or widespread ones. The most
obvious example is the replacement of pati with gopdydti already discussed above. This
occurs throughout the SB, and though pati occurs in unglossed mantras, it does not occur
independently in prose, whereas gopayati is common there.

However, this type of replacement throughout the text is unusual. Instead, it turns out that
a number of systematic replacements are confined to a portion of the text, i.e. to the
Sandilya books VI--X. Thus, the rare root ci ‘perceive’ is replaced by drs ‘see’ 3 times in
Kanda VI. The -dya-formation purdya- replaces forms of pr/pra 5 times in this section, but
elsewhere prnati, dprah are unglossed:

(a) prna — (d)püraya Vl. 5.4.12, VIII. 7.2.6

aprah > apürayati VI. 7.3.10, VII. 5.2.27

apaprivan > apuravati IX. 2.3.17
Moreover, in this section certain forms serve almost as cover-term glosses for a range of

mantra forms. In ımä prohibitives hirhsih ‘do not harm’ glosses four different verbs, each
with a more specific meaning; 3 of the 4 are in Books VI and VII:

(HL.6.4.13- 14 lekhih ^ himsih)
VI.444 abhi socih > himsih
VII. 5.1.8 abhi tapsit > himsit

VII. 5.2.17 abhi mamsthah > himsih

The height of what one must call absurdity is found with the stem dipya- and especially
participle dipyamana- ‘shine, shining’, which appears as gloss 15 times, sometimes to forms
of the root di ‘shine’ to which it was originally related, but to a wide semantic range of
other forms as well, including such divergent ones as mandasva ‘become exhilarated’, ud
iyarsi ‘thou dost raise up’. All 15 examples occur in Books VI--IX:

VI. 3.3.19 (rabhasam) disanam > dipyamänam

VI. 3.3.20 járbhuranah > dipyamanah
VI.4.12 didyatam > dipyamanam
VI.427 didivan > dipyamanah
VI.4.29. SOcasva > dipyasva
VI.4.4.21 sosucanah > dipyamanah
VI.8.1.14 didaya > dipyamanah
VII. 1.1.28 árocathah > ädipyathäh

VII. 3.1.30 ud iyarsi > ud dipyase

175

VII.3.1.31 mándasva — dipyasva
VII.3.1.32 irajyan > dipyamänah
VIII.6.3.20 dévidyutat > dipyamanah
VIIL6.3.21 _ dyotatam > dipyatam
VIIL6.3.21 vibhrajamänah > dipyamanah
IX.2.3.25 didyänah > dipyamanah

These root replacements are not the only feature characteristic of the Sändilya section.
Other peculiarities already discussed are unequally distributed in the text, found much
more commonly here. 3 of the 4 examples of hara- DEDE bhára- occur here (ul, vil,
VII}, VIII!), 6 of the 7 of karóti replacing krnóti (1l, VIS, VII, VIII). In other words,
though krnôti is no longer a freely produced stem anywhere in the SB, it is only in this
section that a concerted effort is made to gloss it with its modern counterpart.

Of course, these strikingly different patterns of glossing in the different sections of the text
do not lead to equally striking new conclusions about the strata of the text, since the dif-
ferent provenience of the Yäjfiavalkya and Sändilya sections has been generally accepted
for well over a century. However, this does lengthen the list, begun by Weber (Ind. Stud.
13: 1873), of significant linguistic differences between the sections, and also makes the
glossing practices within those two different sections seem more self-consistent.

Oddly enough, the type of almost verbatim gloss so common in the SB is relatively infre-
quent in other Brahmana texts,* so that systematic comparison of the development of the
Vedic verbal system is not possible by this method. However, I hope in future to examine
the Srauta Sütra material, which should yield a wealth of data concerning later stages of
the verbal system.

Notes

All passages cited are from the SB (Madhyandina recension).

But note the consistency in, voice in the immediately preceding VII. 2.2.4 sirà yuftjanti kaváyo yugá
ví tanvate prthag iti... té siram ca yunjänti yugäni ca vi tanvate prthak.

For further division of this last type, see my ‘Two problems in the inflection of the Vedic inten-
sive’, MSS 42 (1983) 41--73.

Unfortunately I have had access only to a small portion of the Kanva recension of the SB.

The tenses of the Latin perfect system

Jay H. Jasanoff (Cornell University)

1. One of the most conspicuous features of the Latin verbal system is the contrast between
the three tenses of the infectum, or present system, and those of the perfectum, or perfect
system.’ Thus, a typical primary verb like tango, -ere ‘touch’ forms a present tense proper
(3rd sg. tangit ‘touches’), a future (tanget ‘will touch’) and a preterite or ‘imperfect’
(tangebat ‘was touching’), which together comprise its present system; systematically op-
posed to these are the perfect (£etigit 'has touched"),? future perfect (tetigerit (1st sg. -Ó)
‘will have touched’) and pluperfect (tetigerat ‘had touched’), which constitute the corres-
ponding perfectum. This parallelism extends to the subjunctive, where the opposition be-
tween imperfect (tangeret) and present (tangat — better termed ‘non-preterite’) forms in
the infectum is matched by the contrast between the pluperfect (tetigisser) and perfect
(tetigerit (1st sg. -im)) subjunctives in the perfectum.” Similarly, there are two infinitives:
tangere ‘to touch’ has its structural counterpart in the perfect infinitive tetigisse ‘to have
touched’.

This state of affairs is quite different from that reconstructible for Proto-Indo-European.
Evidence from Indo-Iranian and Greek indicates that the perfect served simply to denote a
state in the parent language, without any overt specification of time; although a tendency
to create distinct tense forms from the perfect stem is well-developed in Greek and Vedic
Sanskrit, the verbal systems of these languages are organized on entirely different lines
from that of Latin.* The Latin system, however, is not wholly isolated. It has an almost
exact counterpart in the Italic dialects, which likewise oppose their present (e.g., Vest.
didet ‘dat’, subj. Umbr. dirsa ‘det’) to a perfect (e.g., Osc. deded ‘dedit’, subj. dadid ‘de-
diderit’) and their future (e.g., Osc. didest ‘dabit’, Umbr. habiest ‘habebit’) to a future
perfect (e.g., Osc. fefacust ‘fecerit’, Umbr. habus ‘habuerit’). It is quite possible that the
poorly represented Osco-Umbrian imperfect (.e.g., Osc. 3rd pl. fufans ‘erant’, patensins
‘panderent’) was paired with a pluperfect, but no example of this tense is attested.

Although the structural similarity of the Latin and Osco-Umbrian systems presumably
reflects a Common Italic feature, the detailed history of the Latin perfect tenses is for the
most part obscure. Outside the perfect indicative itself, the perfectum in Latin is character-
ized by an etymologically opaque element -er- (-is- before consonants), to which are added
the appropriate tense and mood signs — -i- for the future perfect, -a- for the pluperfect, -i-
for the perfect subjunctive, -se- for the pluperfect subjunctive and -se for the perfect in-
finitive. No trace of this formative is found in Osco-Umbrian: here the future perfect is
characterized by a tense sign -us- (Osc. -uz-, Umbr. -ur- before vowels), while the perfect
subjunctive is formed by adding the mood sign -i-, -i- to the unextended perfect stem.

178

These forms, themselves problematic, show no obvious connection with the corresponding
categories in Latin.

2. Given the absence of clear formal counterparts elsewhere, it is hardly surprising that the
majority of attempts to explain the Latin perfect system have been based on internal re-
construction. The results yielded by this method are at least superficially encouraging: each
of the tense and mood signs identified above recurs in a similar function in the infectum,
where each has a well-known Italic or Indo-European etymology. Thus, the historical
thematic vowel of the future perfect appears also in the ordinary futures eri? ‘will be’ and
cantabit ‘will sing’, and ultimately continues the *-e/o- of the Indo-European subjunctive;
the -d- of the pluperfect is identical with the *-d- of imperfects like erat ‘was’ and cantäbat
‘was singing’ and continues a preterital element known also from Celtic (Benveniste 1951:
19; Jasanoff 1983: 75--82); the -i- of the perfect subjunctive is clearly comparable with
the vowel of present subjunctives like sit ‘may be’ and uelit ‘may wish’, which rest on
inherited athematic optatives in *-je-/-i Even the -se- of the pluperfect subjunctive,
though ultimately obscure in Indo-European terms, is inseparable within Latin from the
-sé- (after vowels -ré-) of imperfect subjunctives like esset ‘might be’ and tangeret ‘might
touch’. The formative -er- (-is-) to which these elements are added is clearly the continuant
of an earlier non-alternating *-is-, which is preserved before the *-sé- of the pluperfect sub-
junctive but shows the regular effects of rhotacism and vowel weakening before the vocalic
suffixes of the other categories. The historical inference suggested by the Latin facts, there-
fore, is that forms like tetigerit, tetigerat, tetigerit and tetigisset originated as the ‘short
vowel’ subjunctive, d-preterite, athematic optative and ‘sé-subjunctive’, respectively, of an
enlarged perfect stem in *-is-. This line of reasoning, essentially non-comparative in char-
acter, is accepted by most of the major handbooks of Latin historical grammar (cf. Sommer
1914: 575; Meillet-Vendryes 1953: 265--266; Buck 1937: 297; Leumann-Hofmann 1977:
608ff.).

The source of the enlargement *-is-, however, remains problematic. According to the usual
view, *-is- is of ‘aoristic origin’, and came to be generalized throughout the perfectum from
a nucleus of verbs in which it served as the distinguishing mark of the perfect (< aorist)
stem. Implicit in this analysis is the assumption that alongside the familiar sigmatic aorist in
*-s-, the verbal system of Proto-Into-European also had an aorist in *-is-; the main evidence
for such a category is provided by the is-aorist of Vedic Sanskrit. The Vedic formant in
question, however, is now known to have originated from the addition of *-s- to roots or
stems ending in a laryngeal: 1st sg. dpavisam ‘I purified’ is simply the normal reflex of the
s-aorist “e-péuhx-s-m, while dstambhisam ‘I propped up’ is the sigmatized replacement of
an earlier aorist *e-stembhhy m.? Thus, the -i- of the is-aorist is etymologically not a true
*i- but a vocalized laryngeal, which would have yielded *-4- rather than *-i- in Italic. A
direct connection between the Latin and Sanskrit forms is virtually impossible.

Independent evidence for an aorist morpheme *is- has also been alleged from Latin itself.
It is well-known that the personal endings of the Latin perfect are largely based on those of
the perfect of Proto-Indo-European: thus, the first singular ending -i (tetig?, etc.) continues
the *-a (< *-h2e) of Gk. oia and Ved. véda ‘I know’, augmented by the hic et nunc
particle *i, while the third plural in -ére (tetigére < *-ér-i) is related to the Hittite third
plural ending -er and, more distantly, to the Vedic third plural perfect in -uh (< *-rs). In
the second person, however, the Latin endings are -isti (sg.) and -istis (pl.), which have

179

commonly been regarded as the historically regular perfect forms (cf. especially Ved. 2nd
sg. -tha, Gk. -0a, Hitt. (hi-conj.) -ti < *-tai) preceded by an originally autonomous mor-
pheme *-is-. The validity of this interpretation is questionable. As noted by Cowgill (1965:
172—73) there is considerable comparative evidence for assuming a sigmatic doublet of the
second singular perfect ending in the parent language. Lat. -sti is difficult to separate from
Gk. -o0a, Go. -st (in saisost ‘you sowed’), Hitt. 2nd sg. pret. (ki-conj.) -šta and Toch. B 2nd
sg. pret. -sta (A -st). The source of these variants is unclear, but it is natural to speculate
that *-st(hJa (« *-sthze) arose by resegmentation from cases where a root-final dental in
contact with the *-t- of the original ending produced a sibilant by regular sound change
(*-TT- > *-TsT-).

The conclusion suggests itself, therefore, that the historically correct segmentation of Lat.
«isti is not -is-ti, with an ending -ti accompanied by a tense sign -is-, but -i-stî, with a sig-
matic desinence -sti preceded by a union vowel -i-.© Such an analysis is also favored by the
overall structure of the perfect paradigm. From a synchronic point of view, none of the
perfect endings begins with a consonant — a state of affairs readily intelligible in the first
singular, third singular and third plural, where -i, -it and -ére continue *ai, *-ei[1] and *-éri,
respectively, but more suprising in the second person and in the first plural, where the
regular ending is -imus. The latter termination, a replacement of earlier *-mos, is clearly
secondary. Its origin is commonly traced to the reduplicated perfects dedimus ‘we gave’
and stetimus ‘we stood’, which can theoretically be taken from athematic preforms of the
type "dedhamos and "steth;mos (Sommer 1914: 577—578; Buck 1937: 296; Leumann-
Hofmann 1977: 607). This explanation, however, requires us to separate the -i- of -imus
from that of 2nd sg. -isti and 2nd pl. -istis: the rules of Latin vowel-weakening, which
would regularly have allowed the pre-Latin laryngeal reflex *-a- to develop to -i- in open
syllables, would have taken a sequence like 2nd sg. *dedastai to *dedesti. Phonologically,
the -i- of -isti and -istis can only continue an original St. and is natural to see this vowel
in -imus as well.” From a typological point of view, -isti, -imus and -istis resemble nothing
to closely as the corresponding Greek perfect endings -a«, -auev and -are, the initial ele-
ment of which is likewise an accretion of the post-Indo-European period. While the -i- of
the Latin forms presents far more serious etymological difficulties than its Greek counter-
part, its original status as a union vowel is hardly less clear.?

This finding, though damaging to the traditional view of the extended perfect system, is
eminently consistent with the known tendencies of analogical change. It would be surprising
indeed if Latin had systematically excluded an inherited tense sign *-is- from the first and
third person forms of the perfect, while adding it in the second person even to forms which
were already sigmatic (cf. dixisti ‘you said’, etc.); it would be more remarkable still if such
an element, though nowhere attested outside Latin, had subsequently become an obliga-
tory constituent of the stem from which all the remaining perfectum forms were created in
the post-Italic period. Internal reconstruction, normally an invaluable supplement to the
comparative method, would here seem to have led to an impasse.

3. It may be appropriate at this point in our discussion to recall that the elaboration of the
perfect system was not a development of Latin alone, but of the Italic branch of Indo-
European as a whole. The future perfect and perfect subjunctive of Oscan and Umbrian
bear little overt resemblance to the corresponding categories of Latin, but the formal
disparity between the two groups can easily be exaggerated. In point of fact, the traditional

180

tendency to separate the Osco-Umbrian perfectum from that of Latin is valid enough in the
case of the perfect subjunctive, where forms like Osc. dadid, fefacid ‘fecerit’ and Umbr.
combifiansi ‘nuntiauerit’ are transparently old perfect optatives of a type not found in
Latin.” The situation in the future perfect, however, is of greater potential interest.

A great deal has been written concerning the origin of the Osco-Umbrian future perfect
sign -us-, which in the third singular is followed directly by the personal ending -t (cf. Osc.
fefacust ‘fecerit’, Umbr. dersicust ‘dixerit’) and in the third plural develops regularly to
Osc. -uz-, Umbr. -ur- before the vocalic ending -ent (cf. Osc. angetuzet ‘proposuerint’,
Umbr. dersicurent ‘dixerint’). There is little to recommend the old comparison of these
forms with the Latin ui-perfect, as suggested, e.g., by von Planta (1897: 373ff.) or with the
Indo-European perfect active participle in *-uos-/-ues-/-us- (so Schultze 1887: 272-4;
similarly Buck 1937:173).!° Rather, as seen already by Bartholomae (1887:92), the
point of departure for the creation of the Osco-Umbrian future perfect must have been
the root fu- ‘be’, where both the appearance of the sequence -us- and its eventual re-
analysis as an independent morpheme can be easily explained. The future perfect of
fu- is probably attested in the Umbrian third plural form fefure (for *fefurent), implying
a 3rd sg. *fefust; it is likely that this was the Common Osco-Umbrian form, and that
the absence of reduplication in Osc. fust ‘fuerit’ is secondary. The origin of *fefust is
clear. It is a future perfect of exactly the same type as Gk. redvnkeı ‘will be dead’
and AeAerWeraz ‘will be left’, in which the productive mark of the simple future (-ce/o-) has
been added to the perfect stem to produce a new future tense with stative value. In Osco-
Umbrian the sign of the ordinary future is a bare -s-; *fefust patently shows the addition of
this morpheme to the inherited perfect stem *fefu- (cf. Ved. 3rd sg. babhuva ‘was’, with
secondary -U-).

Once established in *fefust, the spread of -us- to other sigmatic future perfects would have
been favored by two important factors. First, the simple perfect corresponding to *fefust
would presumably have had the Common Osco-Umbrian form *fefed < *fefued (cf. Osc.
3rd pl. fufens, 3rd sg. *fufed, with secondary -u- from fufans ‘erant’), so that the *-u- of
*fefust would early have tended to lose its synchronic status as a constituent of the root.
Equally important, the perfect system of *fu- appeared not only in the free forms *fefust
and *fefed, but in every Osco-Umbrian verb that had an f-perfect — a fact which gains in
significance when it is recalled that the f-perfect is the only one of the suffixed perfect
types to appear in both dialects. The pattern -ed (perf.): -ust (future perfect) would thus
have received support from pairs of forms like Osc. aíkdafed 'decreuit' (?) on the one hand
and Umbr. andirsafust ‘circumtulerit’ on the other; the eventual result was the establish-
ment of a proportion *-fed : *-fust :: *-ed: X, which triggered the creation of future per-
fects like *fefakust (cf. Osc. fefacust), *dedikust (Umbr. dersicust) and benust (Umbr.
benust ‘uenerit’) from perfects of the type *fefaked (cf. Osc. avafaker *dedicauit’), *de-
diked and *bened (cf. Osc. kümbened ‘conuenit’). Such forms apparently supplanted
earlier future perfects in simple *-st (*fefakst, *dedikst, etc.), the replacement of *-s- by
*us- providing a transparent example of a ‘bipartite’ morpheme, in Kurylowicz’ sense,
taking on the distribution and function of a simple one."

The future perfect of the Italic dialects would thus seem to have originated as an s-future
built from the perfect stem. In the discussion that follows, we shall examine the conse-
quences of this result for the analysis of the future perfect in Latin, the development of
which will be seen to have proceeded along strikingly similar lines.

181

4. We have already noted that the traditional interpretation of Lat. tetigerit as the sub-
junctive of a quasi-aorist stem *fetag-is- depends crucially on the untenable assumption of
an is-aorist in the verbal system of Proto-Indo-European. Let us therefore approach the
problem from an Italic, rather than purely Latin, perspective and consider the possibility
that in pre-Latin, as in pre-Osco-Umbrian, the future perfect was originally made by adding
#s- to the perfect stem. A verb like tango would thus have acquired a future perfect stem
*tetag-s-, inflected athematically. The fate of such a stem in Latin would have been closely
bound up with that of the s-future of the infectum, a formation which survives in early
Latin futures like faxo, -if ‘will do’ and capso, -it ‘will take’.!? The faxó-type is thematic,
presumably because the inherited indicative *faksmi, -ti, etc. was prehistorically replaced
by the corresponding subjunctive. It is natural to assume that the same development would
have affected the future perfect, effectively entailing the thematization of *tetag-s- to
*tetag-se/o-. The regular third person singular of this stem would have appeared in Classical
Latin as *tetexit; similarly, from ago ‘I lead’ and dico ‘I say’ the expected future perfects
would have been *exit and *dixit, respectively. These forms, it will be noted, differ from
the attested tetigerit, égerit and dixerit in only one particular: they lack the etymological
*.i- (> -e-) which precedes the *-s- (> -r-) in the historical forms. The problem of explaining
the future perfect thus reduces to the problem of explaining the substitution of the stem-
type *tetag-i-se/o- for the shorter and historically predictable *tetag-se/o-.

An obvious solution emerges from a comparison of our third hypothetical example, "dixit,
with its attested counterpart dixerit. The verb dicó is one of many in Latin with a perfect
that continues an inherited sigmatic aorist (*deik-s-). The addition of the *-s- of the future
perfect to the already sigmatic perfect stem *deiks- would regularly, of course, have had no
overt phonetic affect: *deiks-s- would simply have fallen together with *deiks-, giving the
postulated *dixit ‘will have said’. But dixit, representing older *deik-se-t, was already
current in older Latin as the normal s-future of dicö in the present system.!? The homo-
phony of *dixit, or its prototype *deikset /deiks-se-t/, ‘will have said’ with dixit < *deikset
/deikse-t/ ‘will say’ would hardly have been tolerable in a language where the infectum:
perfectum opposition was as basic a feature of the verbal system as it was in pre-Latin. A
device would clearly have been needed to disambiguate the two forms; it is in this context
that the replacement of the future perfect *deikset, with underlying but unrealized *-ss-,
by *deiksiset, the source of the attested dixerit, may best be understood. We have seen in
section 2 that the union vowel *-i- was prehistorically introduced into the paradigm of the
perfect proper as a means of separating the stem from the consonantal endings. The natural
inference is that this vowel was extended from the perfect indicative to fill a comparable
function in the future perfect. There are several ways in which the generalization of *-i- can
have proceeded. It is possible, for example, that at an early period, when forms like 2nd sg.
*deiks-i-stai were still in free variation with older forms of the type *deiks-stai (i.e., [deik-
stai]), analogy led to the creation of *deiks-i-se-t beside *deiks-se-t ([deikset]). Alterna-
tively, the co-occurrence in early Latin of haplological forms like dixti, misti ‘you sent’,
intelléxti ‘you understood’, etc. beside regular dixisti, misisti, intelléxisti may have led to
the back-formation of *deiksiset from *deikset in its future perfect function, thus making
possible the structurally useful distinction between *deiksiset > dixerit (future perfect)
and *deikset > dixit (future). In either case, the result would have been the substitution of
*-iseo- for *-se/o- as the mark of the future perfect in verbs with s-perfects; from here the
longer and more convenient allomorph would have been free to spread to forms like

182

*tetakset /tetag-se-t/ and *ekset /èg-se-t/, giving *teragiset > tetigerit and *egiset > égerit.
Viewed in this way, the Latin developments would present obvious parallels to those in
Osco-Umbrian, where *-s- was similarly replaced by a variant suffix of the form *-Vs-,
albeit via an entirely different series of analogical changes.

5. An explanation of the perfect subjunctive follows almost mechanically from the fore-
going account of the future perfect. In the infectum, the archaic s-future is associated with
a present subjunctive, represented by forms of the type faxim, -is, -it, ausim (CI would
dare’), 4s, -it, etc.; historically, these simply continue the optative of the athematic forma-
tion presupposed by the type faxo. It is in no way surprising, therefore, that the future of
the perfectum is likewise associated with an old optative, the attested reflexes of which
show the regular renewal of *-s- to *-is- and pattern as perfect subjunctives (cf. tetigerit <
*tetagisit). Sigmatic forms of this type may well have competed for a time with perfect
optatives similar to those of Osco-Umbrian (Osc. fefacid, etc.), but their success at the ex-
pense of the earlier formation was complete by the time of our earliest records.

Unlike the future perfect and perfect subjunctive, which under the above interpretation
can be regarded as analogical transformations of categories inherited from Common Italic,
the remainder of the perfect system is, at least from a formal point of view, of compara-
tively recent origin. The historically correct analysis of tetigerit and tetigerit as *tetag-i-se-t
and *tetag-i-s-i-t, respectively, is no longer valid for the attested stages of Latin; owing to
phonological and other changes, the synchronic form tetigerit bears a much closer surface
resemblance to erit (< *es-e-t), the third person singular future of the copula, than to its
immediate relatives of the type faxit. It is probable that this similarity was exploited in the
creation of the new pluperfect: given the fact that in the infectum erit was opposed to a
preterite (imperfect) erat, it was a simple matter for fetigerat to enter the perfectum as the
preterital counterpart of the future perfect tetigerit. Comparable developments were in all
likelihood responsible for the introduction of the pluperfect subjunctive tetigisset and per-
fect infinitive tetigisse; obvious models were provided by the copula forms esset (impf.
subj.) and esse (pres. inf.).'*

With the establishment in the perfectum of fetigerat, tetigisset and tetigisse, or their pho-
nological antecedents, the metamorphosis of the sequence *-fi)s- from a future sign to an
ancillary mark of the perfect stem would have been complete. In the resulting system of
the historical period, -er-/-is- is a purely formal enlargement with no detectable semantic
function; its history provides a striking illustration of the vicissitudes to which a gram-
matical morpheme — particularly a productive one — may be subject in the course of lin-
guistic change.

Notes

! I would like to thank Alan Nussbaum for numerous comments on an earlier version of this paper.

All errors are naturally my own.

The symmetry of the system, of course, is slightly disturbed by the fact that the perfect may also
have the value of a simple preterite; in this sense it presumably reflects the semantics of the Indo-
European aorist, of which it is in part the formal continuant as well.

As a formal convention in the discussion that follows, the endings of the 3rd singular perfect sub-
junctive and 3rd singular future perfect will be cited as -/f and -it, respectively. In practice, the two
are confused even in Plautine Latin.

183

In particular, these languages assign a central role to the present: aorist opposition, which has no
counterpart in Latin. It should be noted, moreover, that the Vedic pluperfect is not a true past of
the perfect at all, but an ordinary preterite with essentially the same value as the imperfect.

Cf. Narten (1964) s.v. grabhl.. The absence of vrddhi is the surest indication that the -s- is secondary.
Similarly, the second plural in -istis is presumably to be analyzed as -i-stis, with -s- analogically ex-
tended from the second singular.

Note further that the Plautine and Vulgar Latin third plural in -érunt (cf. OFr. distrent, It. dissero,
etc.) can straightforwardly be analyzed as *-i-r[onti]. An active third plural in *-ront is found also in
Tocharian B -re (W. Cowgill, personal communication).

The -a- of the Greek forms presumably originated in the third plural, where the development of
-aT [i] from *-nt was phonologically regular; compare also the *-u- which appears in the plural and
dual endings of the strong preterite in Germanic (Go. pl. -um, -up, -un (< *-nt); du. -u, -uts). Latin
-i-, of course, cannot have arisen in this way; its original locus is not immediately obvious.

A recent account of the Osco-Umbrian perfect subjunctive is given by Lindeman (1982: 303-6).
In the notes to his second edition, however, Buck (1937: 362) abandons this theory for the view
taken here.

Cf. Kurylowicz” first ‘law’ of analogy: “Un morphème bipartite tend à s’assimiler un morphème iso-
fonctionnel consistant en un des deux elements, c.-à-d. le morphéme composé remplace le morphéme
simple." (Kurylowicz 1945—49: 125).

The functional status of these forms, which are only occasionally future perfects, is the subject of a
forthcoming study by Alan Nussbaum.

Such a form is implicit, e.g., in the Plautine 2nd sg. subj. dixis (Capt. 1.2.46).

It is impossible, of course, to determine whether the attested pluperfect indicative and subjunctive
have replaced non-sigmatic predecessors of the type *tetag-d-t and *tetag-sé-t. The group -ss- in the
pluperfect subjunctive and perfect infinitive makes it likely that Indo-European intervocalic *-s- was
still a sibilant (presumably *-z-) at the time of the creation of these forms.

References

Bartholomae, Christian, 1887. “Die Vertretung des altitalischen ss im oskischen (etc.)”. BB 12: 80—92.

Benveniste, Emile, 1951. “Prétérite et optatif en indo-européen”. BSL 47: 11-20.

Buck, Carl Darling, 1937. Comparative Grammar of Greek and Latin (2nd edition) (Chicago: University
of Chicago Press).

Cowgill, Warren, 1965. “Evidence in Greek”, in: Evidence for Laryngeals, edited by W. Winter (London/
The Hague: Mouton), 142-180.

Jasanoff, Jay H., 1984. “The IE ‘ö-preterite’ and Related Forms". IF 88: 54—83.

Kuryłowicz, Jerzy, 1945—1949. “La nature des procès dits ‘analogiques’’’. Acta Linguistica 5: 125ff.

Leumann, Manu/E. Hofmann, 1977. Lateinische Grammatik 1 (München: Beck).

Lindeman, Frederik Otto, 1982. "Notes italiques". BSL 77: 303—308.

Meillet, Antoine/J. Vendryés, 1953. Traité de grammaire comparée des langues classiques (2nd edition)
(Paris: Champion).

Narten, Johanna, 1964. Die sigmatischen Aoriste im Veda (Wiesbaden: Harrassowitz).

Planta, Robert von, 1897. Grammatik der osko-umbrischen Dialekte 2 (Straßburg: Trübner).

Schultze, Wilhelm, 1887. “Das lateinische v-perfectum”. KZ 28: 266--74.

Sommer, Ferdinand, 1914. Handbuch der lateinischen Laut- und Formenlehre (2. Auflage) (Heidelberg:
Winter).

*H3 in Anatolian

Sara Kimball (Philadelphia, Pennsylvania)

One of the most serious problems confronting the linguist who is trying to establish the
phonological history of Hittite and its sister language Cuneiform Luvian (hereafter simply
Luvian) is the syllabary in which these languages are written. The cuneiform syllabary was
a borrowed script which was not completely adapted to the needs of writing either lan-
guage, and it obscures a number of important details. Specifically, there are phonological
distinctions which can be reconstructed for Indo-European (and which might be expected
in Proto-Anatolian) that the cuneiform syllabary apparently does not make. The problem
that concerns us here is whether there was a distinction in Proto-Anatolian or its daughter
languages between initial *h and *h3. Although possible reflexes of both sounds are
written with the same signs in the cuneiform syllabary, evidence from Lycian, which is
written in an alphabet that can make some of the distinctions which are obscured in the
cuneiform syllabary, helps to clear matters up.

It is certainly beyond question that initial *h? becomes the sound that is written with signs
for H plus vowel in Hittite and Luvian, and that the Lycian sound that corresponds to
Hittite and Luvian A is a velar, presumably a velar spirant, written with x (K), g or q.!
Several clear etymologies establish this. One of the best known of these is: Hitt. hant-
“front” (in hanti “separately,” hànza "front," “in front," hantezziya- "first"), Luv.
hantil(i). "first," handawat- "commander": Lyc. 3rd sing. pret. x/izewete "he led" and

xfitawata- “commander,” “basileus”? (IE *haent-, *hant- cf. e.g. Gk. àvrt Lat. ante)?

The problem is with *h3. There are several cases in which a Hittite or Luvian word with
initial h can be compared with words in the other Indo-European languages with an *o that
can perhaps be derived from *h3(e). The best of these equations are: Hittl hara(n)- “eagle”
(and Pal. hara(n)- : Gk. ópveov, öpvis; Hitt. hark- “be destroyed," harni(n)k- destroy,"
harga- “destruction” : O. Ir. orgaid “slays”; Hitt. happar “price,” “deal,” Hitt. (and Luv.?)
happe/inant- “rich” : Lat. opus; Hitt. hissa- “axle, shaft” : Gk. ola&. But there is a certain
amount of evidence suggesting that *h> did not color a neighboring *o and *h3 can only
be reconstructed for any of these words when the o-vocalism of the non-Anatolian lan-
guages is completely anomalous and there is virtually no possibility that it is analogical.
It is extremely difficult, however, entirely to exclude an o-grade, whether original or
analogical, in many cases. The word for "sheep" illustrates the problem. On the basis of
Luvian hawis (H. Luv. hawa/i-) and Lycian xava- we can reconstruct an initial laryngeal for
the Indo-European word. The o-vocalism of Greek di¢ (Hmc. ots), Latin ovis, and O. Ir. oi
suggests that the laryngeal can be *h3 (i.e. *h3ewi-), but *hyowi- is also possible.

186

Even if any of these words can be confidently reconstructed with *h3, it is not clear
whether *h and *h3 remained distinct phonemes in Anatolian or whether they merged in
a single h-like sound. Unfortunately, the writing system offers only ambiguities on this
point, since the reflex of “h and possible reflexes of *h3 are both written with the same
series of signs.

It has also been claimed that *h3 was lost in Anatolian, and there are several plausible look-
ing etymologies which seem to support this. For example: Hitt. aniya- “‘work,” Luv. anniva-
id. (Pal. aniya-): Lat. onus (*h3en-); Hitt. arra- “podex”: Gk. dppos (*h3erso-); ar- “arrive,”
ar- “place onself,” arai- “arise” : Gk. cope, copro, Opvuut, Lat. orior (*haer-); arki-
“testicle” : GK. öpxıs (*hzergh-), and perhaps ariya- “consult an oracle" if this belongs
only with Latin oro, oraculum and not also with Greek dpa, updopat.®

29 Ge

One family of words, however, Hittite happar “business transaction,” “price” and its
cognates in Anatolian and the other Indo-European languages, provides evidence that initial
*h3 was not only retained in Anatolian, it was in fact distinct from *h».

The o-vocalism of the Italic s-stem *opos in Latin opus and the denominatives operor, Osc.
upsannam “‘operandam,” “faciendam” and Umbr. osatu “operator,” "facito" points to
initial *h3 while Sanskrit apas- and Avestan apah- indicate that the s-stem *h3epos, *h3epes-
was Indo-European.” Italic *opos is found beside a related root noun “ops (attested in
the abl. sing. ope, and the nom. pl. opes, and implied by the nom. sing. Opis). *Ops is
found also in the compounds *cóps (acc. sing. copem, abl. sing. copi),? copia, and inops,
and in Old Irish in the adjectives somme “rich” and domme “poor” (*su-/dus-op-smiyo-).
Taken by itself, *ops can continue *h3eps, *h3ops, or *h5ops. If the latter were the
correct reconstruction, then it might be possible that the o-vocalism of Italic *opos was
analogical from *ops.!° The meaning of the two words, however, was not very close in
Italic, and they seem to reflect a semantic split that must have taken place within Indo-
European.

The semantic split is reflected in the other Indo-European languages. On the one hand,
there are words from *h3ep- which refer to **work" or "action" such as Italic *opos, Indo-
Iranian *apas-, Old Norse efni, “material,” efna “work,” Old English efnan id. (Gmc.
*abnia-, denom. *abnjan), Old High German uoben “do, practice, celebrate,” Old Saxon
obian "celebrate" (Gmc. *objan, cf. Mod. Germ. üben), and Old High German uobo,
lantuobo “farmer,” uobari "farmer." On the other hand, many of the Indo-European
languages attest words from *h3ep- which mean “wealth or “abundance,” perhaps as the
products of work or action. In addition to Italic and Celtic *ops, these include, for ex-
ample: Sanskrit apnas- “wealth,” Avestan afnavant- “having wealth,” and perhaps Lithu-
anian @pstas “abundance,” âpstus “wealthy.”*!

The meaning “wealth,” “abundance” is also found in Anatolian, in Hittite (and Luvian?)
happenant- or happinant- “rich,” and in Hittite happe/inéss- “become rich,” and happe/
inahh- “make rich.” A primary verb from *h3ep- is perhaps found in the hapax ha-ap-zi
"be abundant (?)" (KBo XI 34 L. 5). Happar “price,” "deal" (e.g. OH Aa--ap-par KBo
VI 2 II 51, ha-a-ap-pa-ra-az ib. 1l 54, ha-ap-pa-ri- KUB XXIX 29 Vs 117? and the denomi-
natives Aappariya- "hand over" (OH ha-ap-pa-ri-e-nu-un KBo IT 22 L.20), and happaräi-,
happiräi- “sell”'” provide an interesting middle ground between the ideas of “wealth” or
"abundance" and action which results in wealth. The commercial meanings of these words
presumably reflect the fact that commerce was more sophisticated in second millenium
Anatolia than it was in the rest of the Indo-European speaking world.!4

187

A noun spelled ha-appi-ir is found in a Neo-Hittite text. KUB IV 3 + KBo XII 70 (ha-ap-pi-
ir Rs 14, ha-ap-pi-ir-ma Rs 15). Happir clearly means "price" (nu-us-si ha-ap-pi-ir pe-[es-ke]
“give him a price” Rs 14), and seems, therefore, to be related to happar. Another Anatolian
noun from *h3ep- is Hittite happiriya- (NH also happiri-) “city,” “settled place.” This is
usually written with the ideogram URU (e.g. URU-ri KBo VI 2 1 7, Kbo XX 64 Rs 11,
URU-pi-ra-as KUB XXXVI 62 L. 8, URU-az KBo III 22 L. 5, URU-ya-an ib. L. 55, URU-ri-
ya-[an] ib. L. 70), but a dative is spelled out as ha-z-ap-pi-ri in KBo V 6 I 17 (Neo-
Hittite).ó The meaning “city” or “settled place” perhaps indicates that the happiriya- was
originally a market place,!” or the meaning may be derived from the idea of a “worked
over,” “tilled” or “built up” place (cf. e.g. OHG uobo “farmer, colonus”’).

That happar and its cognates did have initial *h3 and that *h3 remained distinct from *h2
is confirmed by the Lycian cognate epirije- “sell.” The word is found in the third person
singular active, epirijeti (TL 111), and it was originally identified by Laroche (1958, 171—
172). Since it is clear that initial *h resulted in a Lycian spirant, the initial vowel of
epirije- suggests that unlike *hy, *h3 was lost in initial position in Lycian. At the very
least, this means that initial *h> and *h3 were distinct phonemes in Anatolian.

It can, however, be argued that the loss of the initial laryngeal in epirije- is due to a second-
ary sound change, perhaps one peculiar to Lycian. Tischler (1980, 510 n. 52), who notes
the importance of epirije-, prefers this explanation, and Oettinger (1979, 353 n. 200)
suggests that epirije- has a prothetic vowel. There is, however, no evidence to suggest that
prothetic vowels developed before *p in Lycian (cf. pri-, prije- and pere from *pr(e)i- and
*pro!*). More importantly, several considerations indicate that the root vowel of the noun
from which epirije- was derived had the full-grade.

The stem of epirije-, of course, recalls the stems of happirai- “do business,” happiriya-
“city,” “settled place” and the hapax happir “price.” For the vowel of the initial syllable,
the correspondence Hittite a : Lycian e is well known. In other examples the Hittite and
Lycian vowels are from a full-grade *e, *o or *a (cf. e.g. Hitt. Kkweras, *kweran : Lyc. tere,
terñ «. *kVer-; Hitt. pat- (patà-), Luv. pat- : Lyc. ped- « *pod- or *péd-; and Hitt. appan,
Luv. appan, Lyc. epfi « *apm).!°

The resemblance between epirije- and happiräi-, however, is probably more or less coin-
cidental since happirai- is clearly more recent than happarai-. It is first found in Middle
Hittite texts (e.g. ha-ap-pi-ra-[si] KUB XXIII 77 + L. 64; ha-ap-pi-ra-at-te-ni KUB XXIII 72
Rs 58) while happarai- is attested in the Old Hittite copy of the first part of the Law Code
(3 sg. pres. ha-ap-pa-ra-iz-zi KBo XIX 1 II 17;3 sg. pret. ha-ap-pa-ra-a-it KBo VI 21152).
This suggests that happirai- is a replacement for Old Hittite happarai- The most likely
source for the vowel of the second syllable of happirai- is contamination with happiriya-
“city,” “settled place.”

Although Oettinger (1980, 147-149) reconstructs a nominative *hop-ér and claims that
this is directly attested in happir, the hapax happir probably also does not have anything
directly to do with epirije-. The text in which happir is found, KUB IV 3 + KBo 70, is not
particularly old, and there is little reason to believe that it preserves archaisms. In the
Inhaltsübersicht for KBo XII 70, Otten suggests that the text, which is a copy of Akkadian
omens, is a school exercise. It seems very unlikely, therefore, that happir represents an
archaic form of the word for “price,” or that it reveals very much about the Anatolian or
Indo-European paradigm of happar. Instead, happir can probably be considered a mistake,

188

perhaps based on a folk or spelling etymology to happiriya- “city, settled place’ and
happirai- *do business."??

Happiriya- itself appears to continue something original, and it (along with happiriya-
"hand over") is probably the word that is best compared with epirije-. In phonological
terms, the vowels of the second syllables of epirije- and happiriya- can continue Indo-Euro-
pean *i since a correspondence Hittite i : Lycian i is found in Hittite *melit (Post OH
milit)*’ : Lycian melite (uedurn) < Indo-European *melit. A Proto-Anatolian *hap-ir (Indo-
European *h3ep-ir), however, would be very unlikely in morphological terms. The vowel of
the second syllable cannot continue Indo-European “e if a correspondence between Hittite
e and Lycian e (< *a) is found in Hittite kKweras, *kweran “field” (kwer- “cut”) : Lycian
tere, teri “division of an army." Unaccented *e would seem to be ruled out by the corre-
spondence Old Hittite i : Luvian a in Old Hittite nepis : Luvian tapassa- (Indo-European
*nebhes-), since this implies that unaccented *e became *a (> Lyc. e) in the Luvian branch
of Anatolian. The vowel of the second syllable can, however, continue Indo-European *e.
This is indicated by the correspondence Old Hittite i : Luvian i ori : Lycian i seen in Old
Hittite kissar : Luvian issari-, isri- : Lycian izri-. The Anatolian words come from an earlier
*ghés-r-, since *ghes-r- should result in Luvian *iyass(a)r- with a rather than i For the
reflex of *ghe- in Luvian, cf. fivammi- from the locative stem *dhgh-em- beside nomina-
tive-accusative *dhegh-om in Hittite tekan.

We can probably reconstruct Proto-Anatolian *hzap-ér as the noun from which epirije-
and Aappiriya are derived, since the plene writing in NH ha-a-ap-pi-ri points to a full-grade
accented root. Although a double full-grade, *h3ap-er (Indo-European *h3ep-er) would be
unlikely morphologically, the full-grade root *h3ap- can be analogical from the related
*h3ap-ar (Hitt. happer, Indo-European *h3e/op-r), which is surely not an Anatolian inno-
vation. Anatolian *h3ap-ér then, would represent an earlier *h3p-er. In functional terms,
*h3pér could well be a collective beside singular *h3e/op-r.”

The relationship of the stem happe/in(a)- to happar and *h3apér is not entirely clear.
In the first place, the vowel of the second syllable of happe/in(a)- is written with the
ambiguous sign P/ (i.e. ha-ap-Pl-na-), and it cannot, therefore, be determined whether
happe/inant-, happe/iness- and happe/inahh- had a stem with e ori. An Old Hittite happen-
should continue an earlier *h3(e)p-én- (cf. wifen- e.g. in NH dat.-loc. sing. w-i-te-e-ni KBo
V 2 Ii 12 from *wed-én- for *ud-én- beside nom.-acc. watar < *wod-1). If ha-ap-pe/i-na-
can be read as happen(a)-, then it could continue the oblique or locative stem of *h3e/
op. or *hgp-ér, beside *h3ep-n- (for *h3p-n-) in Indo-Iranian *apnas- and Germanic
*abnia-. However, if the vowel of the second syllable was Old Hittite i (i.e. happin-) then
the word could not be derived from an r/n-stem.

Unfortunately, it does not seem to be possible to determine whether happe/inant- is
Luvian as well as Hittite. The word is found once with the *Glossenkeil" (KUB XVII 24
II 17, NH), but this does not by any means guarantee that happe/in(a)- is a genuine Luvian
word." An accusative plural with the Luvian ending -anza, ha-ap-pe/i-na-at-ta-an-za is
found in KUB XXXVI 49 IV 9, a text which is not otherwise especially Luvian looking.??
This does not, however, provide any more certain evidence that the word is really Luvian,
since the Luvian accusative plural endings are sometimes found with stems that are clearly
Hittite (cf. e.g. i-da-a-la-u-wa-an-za KUB XXIX 7 II 29 with Hittite idalu- instead of Luvian
addu wal-.

189

This is not simply hairsplitting. It is otherwise well known that *e should result in Luvian
a, and this indicates that if happe/inant- is a genuine Luvian word, then it cannot continue
*h3(e)p-en- from an r/n-stem. However, if happe/inant- is Luvian, then the vowel of its
second syllable can be from *é, *i or *ei. Since *h3(e)p-én- would not be likely in mor-
phological terms, then happe/inant- should be from *h3(e)p-i-no- or *h3(e)p-ei-no. In this
case, Luvian happinant-, and most likely Hittite happinant-, would be from a derived ad-
jective in *-i-no- or *-ei-no- “pertaining to wealth,” “wealthy” of the type seen in Sanskrit
devinah and Latin divinus.

To sum up: It is well known that initial *h> was preserved in Hittite and in Luvian-Lycian.
Lycian epirije- “sell” and Hittite happiriya- “city,” “settled place,” which can be derived
from an Anatolian *h3ap-er (for *h3p-er) furnish evidence for the history of *h3. The
correspondence Hittite A : Lycian zero in these words shows that initial "ho and *h3
remained distinct in Proto-Anatolian and in the Luvian-Lycian branch of Anatolian.
Although direct evidence for a distinction between *hz and *hg in Hittite is not available
because of the limitations of the writing system and it is possible that they had merged in a
single sound written with the signs for H plus vowel, this is not very likely since Hittite is
usually more conservative than Luvian-Lycian.

However, since *h3 was preserved in Hittite, whether or not it was distinct from “ho,
etymologies which depend on a loss of initial *h3 have to be revised. In many cases, this
is relatively simple. For example, arra- and Greek dppoc can be derived from *hjorso-,
ar-, ar-, arai- while Greek @pope, pro and 6pruu and Latin orior can be from *her-, and
arki- and Greek Öpxıs can continue *h orgh-.% Ariya- “consult an oracle” can be from
*hr-ye/o- and related to Homeric epew, eipoyaı “ask” (*hyr-eu-, *hjr-u-) rather than to
Latin oro, oraculum.

Finally, the equation Hittite h : Lycian zero has further implications for our understanding
of Indo-European phonology. The initial consonant of Lycian xava and Luvian Aiawis can
now be seen clearly to continue *h» instead of *h5. Therefore, the o-vocalism of Greek
dis, Latin ovis and Old Irish oi points to an o-grade root, and this in turn provides an
additional piece of solid evidence that *h^ did not color a following *o.?*

Notes
! For the writing of Lycian reflexes of laryngeals and for the transciption of Lycian k as x, see Houwink
Ten Cate (1961, 111—112) and SevoroSkin (1968, 168).

See Heubeck (1979, 247-260) on the meaning of xfitawata.

The correspondence Hitt. h, hh : Lyc. x, g orq is, of course, also found in medial position, e.g. Hitt.
tarh-, tarhu-, taruh- “conquer, prevail over,” Luv. DTarhunt- “the Storm God" : Lyc. trqgas, trqqis,
traqñt id. (IE *trh9(u)-, or Luv. 1 sg. pret. ending -ha (Hitt. -hhun) : Lyc. -xa, -g2 (Anatol. *-h)a).
The details of this, however, are much less clear than the development in initial position. It is also
not clear whether the different Lycian spellings reflect a split, but see, for example Sevoroëkin
(1968, 168) and the very important findings of Davies (1982/3) on lenition in Luvian.

Evidence for a lack of coloring of the vowel in the sequences *h20 and *oh) is summarized by
Beekes (1972). Alternative views are those of Ruijgh (1970/71), Kortlandt (1980, 127—129) and
Lindeman (1981, 23--31).

For example, Oettinger (1979, 367 n. 212) reconstructs *hoowi-. For *h5owi- as a possibility, see
Beekes (1972, 129 with n. 14) and Rix (1976, 69). *h3ewi- is reconstructed by, for example, Linde-
man (1970, 35 and 1982, 27), Puhvel (1965, 88 and 92), Hamp (1978, 67 n. 7), Rix (1976, 46 and
and 146), Schindler (1969, 153 n. 60) and Tischler (1980, 502 n. 36).

190

13
14
15

17
18
19

21
22

For *h3 in aniya- see, for example, Oettinger (1979, 345) and for *h3 in är- and ar-, see Rix (1965,
28-29 n. 12 and 1969, 92-93), Eichner (1979, 84) and Oettinger (1979, 369, 404 n. 13, 479, 523
and 546). Initial *h3 is reconstructed for UZU ärra- and for arki- by Tischler (1980, 502). Eichner
(1973, 81-82 and 1979, 87) and Oettinger (1979, 43) reconstruct äkfk)- “die” with #h3 (pf.
*h4e-h 4-ok- : *h3e-h3k-) and compare Gk. wxus, Skt. asu and Lat. ocior, but this etymology is not
very convincing semantically. Ariya- was connected with Lat. dro by Pedersen (1934, 47~--48) and
Tischler (1977, 56—57 with Pokorny, 781) adds Russ. orith, Skt. aryate and Gk. dpa and apdouat.
However, if the or- of Ord and orith does continue "haer, then ápa and apaopaı do not belong here,
since the a of these words should be from *h»(e)r-.

Apas- is found beside a noun äpas- “religious work” (2 Vedic examples: see Grassmann (1955, 174)).
The 2 of apas- recalls the *ó of Osc. 3 sg. perf. upsed "fecit," 3 pl. uupsens "fécérunt" and OHG
uoben, OS. obian etc.

Ernout-Meillet (1959, 463—464).

Ernout-Meillet (1959, 464).

This is suggested by, for example, Ernout-Meillet (1959, 466).

Szemerényi (1964, 146 with n. 3) prefers to derive the Lithuanian words from ap/i&) around" plus
*stehy- “stand.” Greek äpevoc, a word which has been compared with kappe/inant- and apnas- etc.,
is most likely a loanword from an Anatolian language. See Szemerényi (1964, 147) and Tischler
(1977, 165) with references.

For the analysis of ha-ap-pa-ri-us in KUB XXIX 29 Vs 11 as dat. sg. happari plus the acc. sg. enclitic
pronoun -us, see Neu (1974, 107).

For the meanings of happariya- and happaräi-/happiräi-, see Oettinger (1979, 352-353).

See Neu (1980) on various terms associated with commerce in Hittite Anatolia.

Happiran in KUB XIII 2 IV 16, which Neu (1974, 108 n. 244) lists as a form of häppiriya- “city,”
should probably be a participle from happiriya- “give over” = “sell” (i.e. happirliy lan). The passage
in which it occurs reads: IV 15 na-as-ma-za da-a-an ku-is-ki ku-it-ki har-zi (16) na-as-ma-za ha-ap-pi-
ra-an ku-is-ki ku-it-ki har-zi “oder jemand etwas (weg)genommen hat (16) oder jemand etwas ver-
kauft hat” (translation, von Schuler, 1957 p. 51).

For the analysis of happiriya- as an -iya- (*-iyo-) stem from which an i-stem häppiri- was back-formed,
see Neu (1974, 108—109).

See Neu (1974, 108) for this meaning.

For Lyc. pere, and Hitt. parà from *pró, see Kimball (1983, 766—777).

For Lyc. fere : Hitt. kweras, see Pedersen (1945, 47—50). According to Laroche (1967, 57—59) tere
tere is a different word, "everywhere." Hittite pat- (patà-) should, of course, continue *pòd-, and
Luvian påt- and Lycian ped- can be from *pòd- or *pèd-. For the long vowel, see Kimball (1983,
864). For Hittite and Luvian äppan and Lycian epf from *apm, see Kimball (1983, 138 and 766).

It is very unlikely that the spelling ha-ap-pi-ir (or ha-ap-pe-er) indicated that the sign PAR can
sometimes be read as PI-IR (PE-ER) as is claimed by Oettinger (1980, 149). Although a few ex-
amples of an apparent alternation PAR ~PI-IR ot PE-ER are found, they are by no means con-
clusive, and they may admit of other explanations. See Kimball (1983, 131).

For Post Old Hittite mi-li-it for Old Hittite *mélit, see Kimball (1983, 13).

Another example of a collective in *-ér might be Hittite hasduér “twigs, brush”; see Schindler (1970,
9).

Happe/in(a)- has been considered either a form derived from the oblique stem of an r/n-stem of
which the nom.-acc. is represented by happar, or a form in *-e-no-. See the references provided by
Tischler (1977, 165 —166).

See Güterbock (1956, 131) on the unreliability of the Glossenkeil here.

Hoffner-Güterbock (1980, 21) consider KUB XXXVI 49 an Old Hittite text with the old ductus, but
linguistically, the text is not especially archaic looking.

For *h orso- see, for example, Eichner (1974, 54). Initial *h, in @r- etc. and in Skt. ara and Gk. is
indicated by the € of £pero «»pu1)0 (Hesych.) and the second element of Aa-ép-rn«. For *h jor&h-
in arki- and öpxıs see Watkins (1975, 12) and Beekes (1969) — with reservations.

See Kimball (1983, 632 n. 111) for this etymology.

Another example of Than resulting in 0 rather than @ would be Gk. doreov, Lat. os : Hitt. hastät.,
Luv. hässa- etc. The Greek and Latin words should continue an o-grade *h)ost- since the a of Welsh
asgwrn (*hest-) points clearly to *hz (Lionel Joseph, personal communication).

191

Bibliography

Beekes, R.S.P., 1969. The Development of the Proto-Indo-European Laryngeals in Greek. (The Hague:
Mouton).

Beckes, R.S.P., 1972. “H 0.” Sprache 18, 117-131.

Cowgill, W., 1965. ‘Evidence in Greek." Evidence for Laryngeals. Ed. W. Winter. (The Hague: Mouton),
142-180.

Davies, A.M., 1982/3. “Dentals, Rhotacism and Verbal Endings in the Luvian Languages.” KZ 96/2:
245-270.

Eichner, H., 1973. “Die Etymologie von heth. Mehur." MSS 31: 53-108.

Eichner, H., 1975. “Die Vorgeschichte des hethitischen Verbalsystems.” Flexion und Wortbildung. Ak-
ten der V. Fachtagung der Indogermanischen Gesellschaft, Regensburg 9--11 September 1973. Ed.

. H. Rix. (Wiesbaden: Reichert), 71—103.

Ernout, A. and A. Meillet, 1959. Dictionnaire étymologique de la langue latine. Histoire des mots.
Quatriéme édition, revue, corrigée et augmentée d'un index. (Paris: Klincksieck).

Goetze, A. and H. Pedersen, 1934. MurSili$ Sprachláhmung. Copenhagen. Det Kongelike Danske Vi-
denskabernes Selskab, Historisk-filologisk Meddelelser 19/1.

Grassmann, H., 1955. Wörterbuch zum Rig-Veda 3. unveränderte Auflage. (Wiesbaden: Harrassowitz).

Güterbock, H., 1956. “Notes on Luwian Studies.” (Review of B. Rosenkranz, Beiträge zur Erforschung
des Luwischen.) Orientalia 25: 113—140.

Hamp, E., 1978. “On Greek Prothetic Vowels.” MSS 37: 59—64.

Heubeck, A., 1979. “Lyk. xñtawata-.” Studia Mediterranea Piero Meriggi Dicata, Vol. 2. Ed. O. Carruba.
(Pavia: Aurora), 247 —260.

Hoffner, H. and H.G. Güterbock, 1980. The Hittite Dictionary of the Oriental Institute of the Uni-
versity of Chicago, Vol. 3, Fasc. 1. (Chicago: The Oriental Institute of the University of Chicago).
Houwink Ten Cate, Ph.J., 1961. The Luvian Population Groups of Lycia and Cilicia Aspera During the

| Hellenistic Period. (Leiden: Brill).

Kimball, S.E., 1983. “Hittite Plene Writing.” Dissertation (University of Pennsylvania).

Kortlandt, F., 1980. “H20 and oh2.” Lingua Posnaniensis 23: 27—29.

Laroche, E., 1957/8. “Comparison du Louvite et du Lycien I.” BSL 53: 159—197.

Laroche, E., 1967. **Comparison du Louvite et du Lycien IV." BSL 62: 46—66.

Lindemann, F.O. Einführung in die Laryngeal-Theorie. Unter Mitwirkung von C.H. Borgstróm. (Berlin:
de Gruyter).

Lindeman, F.O., 1982. The Triple Representation of Schwa in Greek and Some Related Problems of
Indo-European Phonoiogy. Oslo. Institutet for Sammenlignende Kulturforskning. Serie 13: Skrifter,
65.

Neu, E., 1974. Der Anitta Text. StBot 17. (Wiesbaden: Harrassowitz).

Neu, E., 1980. “Die hethitischen Verben des Kaufens und Verkaufens.” WO 11: 79-89.

Oettinger, N., 1979. Die Stammbildung des hethitischen Verbums. Erlanger Beiträge zur Sprach- und
Kunstwissenschaft 64. (Erlangen: H.C. Nurnberg).

Oettinger, N., 1981. “Hethitisch ganenant- ‘gebugt, durstig.’ Ein Beitrag zur Heteroklise.” MSS 40:
143—153.

Pedersen, H., 1945. Lykisch und Hittitisch. Copenhagen: Det Kongelike Danske Videnskabernes Selskab.
Historisk-filologiske Meddelelser 30/4.

Puhvel, J., 1965. “Evidence in Anatolian.” Evidence for Laryngeals. Ed. W. Winter. (The Hague: Mou-
ton), 79-92.

Rix, H., 1965. “Hom. dpwpera: und die Verben 6pvuu und öpww.” IF 170: 25—49.

Rix, H., 1969. "Anlautender Laryngal vor Sonant im Griechischen.” MSS 27: 79-110.

Rix, H., 1976. Historische Grammatik des Griechischen: Laut- und Formenlehre. (Darmstadt: Wissen-
schaftliche Buchgesellschaft).

Ruijgh, C.J., 1970/71. Review of Beekes, The Development of the Proto-Indo-European Laryngeals in
Greek. Lingua 26: 181-198.

Schindler, J., 1969. “Die idg. Wörter für Vogel und Ei.” Sprache 15: 144-167.

Schindler, J., 1974. “L’Apophonie des themes indo-europeens en -r/n.” BLS 70: 1--10.

von Schuler, E., 1957. Hethitische Dienstanweisungen für höhere Hof- und Staatsbeamte. Ein Beitrag
zum antiken Recht Kleinasiens. (Graz: Archiv für Orientforschung), Beiheft 10.

192

Sevoroïkin, V., 1969. “Zur Entstehung und Entwicklung der kleinasiatischen Buchstaben-Schriften.”
Kadmos 7/2: 150-172.

Szemerenyi, O., 1964. Syncope in Greek and Indo-European and the Nature of the Indo-European
Accent. (Naples: Instituto Universitario di Napoli).

Tischler, J., 1977. Hethitisches etymologisches Glossar. Mit Beiträgen von Günter Neumann. Lieferung
1. Innsbruck: Innsbrucker Beiträge zur Sprachwissenschaft 20.

Tischler, J., 1980. “Hethitisch und die Rekonstruktion des indogermanischen Phoneminventars.” Laut-
geschichte und Etymologie. Akten der VI. Fachtagung der Indogermanischen Gesellschaft Wien,
24.-29. September 1978. Ed. M. Mayrhofer, M. Peters and O.E. Pfeiffer. (Wiesbaden: Reichert),
495-522.

Watkins, C., 1975. “La famille indo-européenne de grec. öpxıs: linguistique, poétique et mythologie."
BSL 70: 11-26.

The two senses of the term ‘anaphora’ and their functional unity:
evidence from the Rigveda n

Jared S. Klein (University of Georgia)

O. The term ‘anaphora’, when applied to linguistic contexts, has from classical antiquity
been used in two distinct senses. The earliest securely datable employment of the term in
linguistic value given by Liddell-Scott-Jones in their Greek-English Lexicon is the adjectival
form avayopiKos appearing in the Téxyn of Dionysius Thrax (second century B.C.). In
listing various types of nouns (including proper noun, epithet, homonym, synonym, eponym,
ethnic, and pronoun!) he mentions the “anaphoric [sc. noun] ..., which is also called
homoiomatic [i.e. denoting resemblance] and deictic and antapodotic [i.e. correlative]"
(’Avapopırdv [sc. övoua]. ... ‚ö Kal QuowopariK ov Kat Se TiK Ov kac üvramoborucóv kaAetrat
Uhlig 1883: 40). He next characterizes the anaphoric noun as signifying likeness (70
duoiwow onuaivor), giving as examples the Greek pronouns TowovTos ‘such (a)’, TooovTos
‘so great, so many, so much, etc.’, and rnAwkovros ‘so great, so old’. This same usage is
found three centuries later in the second century A.D. treatise ILepc 'Avrcovuuuas of Apol-
lonius Dyscolus. The latter, however, distinguishes between deictic and anaphoric roles,
noting that "every pronoun is either deictic or anaphoric” (IIaoa aàvrcvuyua N deiktırn
éoTw 1; àvapopuct) Schneider 1878: 9), but that “first and second person pronouns are only
deictic, third person pronouns both deictic and anaphoric” (ai kata mpwrov kar Sevtepov
Hd Sen Tikat, ai karà TÒ Tpurov Kai BekTKal Kai avapopixar Schneider 1878: 9-10).
For both of these important ancient grammarians, therefore, ‘anaphora’ meant resemblance
of reference or referential identity treated as a function of pronouns.

Beside the usage of ‘anaphora’ just described within the field of grammar, the term was
employed in a different sense within the separate tradition of classical rhetoric. Here it
meant not referential identity, but rather repetition of a word, especially at the beginning
of successive clauses. The earliest reference to ‘anaphora’ (this time the noun) in this sense
is found in the treatise Hepi 'Epunveas ‘On Style’ of one Demetrius, dating from the first
century, either B.C. or A.D.” Demetrius states that Sappho “sometimes makes graceful
use ... of anaphora” (Xapıevriferau 6é more ... éé avapopas Roberts 1902: 138), citing
the passage ‘Eotepe, mavta pépes, pépers biv, pépets atya, pépes uarépr maida ‘O Hes-
perus, thou bringest all things (to rest). Thou bringest the sheep, thou bringest the goat
(to the fold), thou bringest the child to the mother’. He then notes that “the charm of this
passage lies in the [repetitious] statement of the [word] ‘pépes’, which repeats [i.e. refers
repeatedly to] the same thing” (n xapıs Eoriw Ek Ts Mékews TAS ‘pépeic’ Ent TO avrò
avayepouévns Roberts 1902: 138). At a later point in the same essay he ascribes a power-
ful effect to “what is called anaphora" (éx eme àvapopac kaXovuévnc), citing the lines of

194 .

Aeschines, ert aavróv kaAeíc, ert roUc vóuouc kaXeic, Emi Tv ômuokpariar kaAetc (1902:
190) ‘Against yourself you summon him, against the laws you summon him, against the
democracy you summon him’. Finally, he notes that this aforementioned figure is three-
fold: “It is epanaphora ... on account of the repetition of the same word at the beginning
of each clause, and asyndeton, because it is uttered without conjunctions, and homoiote-
leuton, on account of the recurrent termination KaAetc" (1902: 190).

We are left, then, with two clearly related but distinct senses of the term ‘anaphora’, and we
can speak of anaphora qua repetition and anaphora qua coreference, with the latter to be
understood as repetition of reference with or without lexical repetition. Although both of
these senses are widely used at the present time, the only sense of ‘anaphora’ in use in
English until the twentieth century was the rhetorical one (anaphora qua repetition). For
instance, the Oxford English Dictionary lists as its only definition of the term *'the repeti-
tion of the same word or phrase in several successive clauses" (Murray, et al 1933: 307).
The first reference to anaphora qua coreference is found in the 1972 supplement to the
Oxford English Dictionary, where it is defined as "the use of a word which refers to, or is a
substitute for, a preceding word or group of words" (Burchfield 1972: 82). The citation
given is from Bloomfield (1933: 266): “The word one .. . replacesa with anaphora of the
noun when no other modifier is present (Here are some apples; take one)". It is not hard to
determine that Bloomfield took the term in this sense from Jespersen (1914: 247), who is
cited in Burchfield (1972) as having been the first to employ the adjective ‘anaphoric’,
applying it also to the term ‘one’ “if it refers to some word already mentioned" (Burchfield
1972: 82). Jespersen may have in turn gotten the term in this sense from Windisch (1869),
who employed it with reference to the citation from Apollonius Dyscolus provided above.?
Over the past twenty years or so ‘anaphora’ in the sense ‘coreference with what precedes’
has become widely used by linguists. At the same time, however, it continues to be em-
ployed by classical philologists in the sense ‘lexical repetition’. Specialists in each of these
areas may not always be aware of the different senses in which ‘anaphora’ is used. It there-
fore seems fitting, in a volume honoring Professor Henry M. Hoenigswald, whose scholar-
ship encompasses both classical philology and linguistics, to point out this difference. How-
ever, in the remainder of this paper we shall demonstrate, using data from the Rigveda, that
the two senses of ‘anaphora’, although different on a purely descriptive level, are equivalent
on the level of discourse structure.

1. That anaphora qua coreference and anaphora qua repetition are closely related in func-
tion is hardly a new idea. As noted by Halliday-McIntosh-Strevens (1964: 248), grammatical
anaphora, grammatical substitution, and lexical anaphora are all features of textual cohesion.
Of these, the first is anaphora qua coreference, the third anaphora qua repetition, and the
second is described as “the use of ‘do’ and ‘one’ ..., as in ‘T might do’ and ‘a big one’”’.
Again, in discussing Indo-European discourse structure Bader (1979: 129) states that “l’ana-
phore a deux définitions (lien avec ce qui précéde; répétition)". These two subtypes of ana-
phora intersect in those instances where the repeated lexical element is itself a pronoun co-
referential with its predecessor, e.g. RV 1.100.13c tám sacante sanáyas tám dhänani ‘That
one do the attainments accompany, that one the prizes'. The role of anaphora within this
passage is inherently conjunctive; it at once demarcates the onset of the individual clauses
and binds them together, through both repetition and coreference, in the absence of overt
conjunction (cf. Demetrius’ statement linking the figures anaphora, asyndeton, and ho-

195

moioteleuton). In most instances in the Rigveda coreferential sequences do not involve
lexical repetition. This is especially true of the correlative structure yd- ... sd/tä-, which
represents a fundamental discourse strategy, as in 1. 164.23d ya it tád vidus té amrtatvám
anasuh ‘The ones who know that, those have attained immortality’. It will be evident,
however, that considered from the point of view of discourse function, the yd- . .. sá/tá-
structure of this latter passage is precisely equivalent to the tam... tam structure of I. 100.
13c. In each case the coreferential terms demarcate and hence bridge the asyndetic gap
between two adjacent clauses. Indeed, iteratively anaphoric terms need not be inherently
referential. Preverbs, negative particles, other adverbials and verbs may be so used as well
with identical effect, e.g. VII. 86.5ab dva drugdháni pítryà srjà no | (á]va yá vayám cakrma
tanubhih ‘Release from us the sins of our fathers (and) those which we have committed by
ourselves’, where the repetition of the preverb ava works in concert with the metrical struc-
ture both to demarcate and bind together the members of the compound object of áva srj
‘release’.

The contention that anaphora qua coreference and anaphora qua repetition are really one
and the same on the level of discourse structure would be substantially confirmed if we
could find in natural language some linguistic form that signaled or was associated with
anaphora in both of these senses. In fact, the Rigvedic particle u is such a form. As I have
shown (Klein 1978a, 1985 part 2: 6—62), this particle finds its fundamental employment
within coreferential sequences of the type yá- ... sá/tá-, noun . . . sd/td-, and sd/td- . . .
Sá/tá-, occurring in second or anaphoric position within these sequences and, through an
inversion, in initial or annunciatory (more specifically, cataphoric*) position. Examples are
the following:

VIIL. 21.9ab yó nah .. . / prá vásya äninaya tam u va stuse The one who has led us forth
unto what is better, that one do I praise for you.’

VIII. 96.6ab tam u stavama ya ima jajana / visvà jitany dvariny asmat ‘That one shall we
praise, who has begotten all these races lower than he.’

VII. 44.6c agním ile sá u sravat ‘Agni do I summon. (And) he shall listen.’

1.35.3cd tám ü sücim sücayo didivansam /apàm ndpätam päri tasthur äpah “That one,
the pure, shining Apam Napät, have the pure waters surrounded.”

IV.8.4ab sd höta sed u dütyam / cikitvañ antdr iyate ‘That one is the priest. (And) that
one, understanding, goes between (heaven and earth) on his embassy.’

II.20.4a tám u stusa indram tám grnise ‘That one, Indra, do I praise, (and) that one do I
sing.’

As we have noted above, the sequence tám ... tám is anaphoric in both senses, and when
accompanied by u following the second occurrence of the pronoun, becomes the channel
whereby the particle expands its range of occurrence from referentially anaphoric to
iteratively anaphoric sequences, subsequently occurring with repeated nonreferential terms,
such as preverbs and verbs, as in 1.91.18a sam te pdyansi sam u yantu vajah ‘Let the milks
unite for thee and the booties unite’ and I1.35.15ab dydisam agne suksitim jdnaya |
dyänsam u maghdvadbhyah suvrktim ‘O Agni, I have extended secure dwelling to the folk,
and I have extended a hymn to the liberal ones’. Again, such sequences are originally
asyndetic, but the particle is susceptible to reanalysis as a conjunction, whence its further
use as a sentential conjunction within noniterative, noncoreferential sequences, e.g. V. 83.

196

10a dvarsir varsám üd u sü grbhäya “Thou hast rained the rain. And do thou make it cease’
and IN. 5.3cd à haryato yajatah sanv asthad / ibhùd u vipro hävyo matinim ‘The one
worth longing for, worthy of worship has climbed upon the back (of the altar). And the
poet, the one to be called upon with thoughts, has come into existence’.®

Once we perceive the role of anaphora as an exponent of discourse cohesion, we can more
easily understand the seemingly pleonastic usage of the sd/td- pronoun in initial position
within clauses having second person verbs. We note the following examples:

III. 19.5ac ydt tvà hôtäram andjan .../... devah // sé tvém no agne (ajvitéhá bodhi
‘Since the heavenly ones have anointed thee Hotar ... As such, do thou become an aider
for us here, O Agni.’

VI.21.7cd.8ab táva pratnéna yujyena sakhyä | väjrena dhrsno dpa tá nudasva /// sa mi
srudhindra nütanasya | brahmanyatáh . . . "Thrust off those (demonic apparitions), O bold
one, with thy ancient ally, thy friend, the cudgel. / Do thou hearken, O Indra, unto the
current utterer of the formulation . . .'

V1. 64.6 sugötd te supdthä pärvatesv | aväte apás tarasi svabhano // si na d vaha . . . / ravim

divo duhitah ... ‘Good ways and good paths are thine among the mountains. (Even) in a

place without wind thou crossest the waters, O self-radiant one. As such, convey hither to
. wealth, O daughter of heaven... .’

In each of the above passages involving, respectively, an overt second person pronoun (sd
tvam), an archaic second person pronoun functioning synchronically as a particle (sd fu, cf.
Klein 1982: 9—11), or no overt second person pronoun (så na d vaha), the synchronic func-
tion of sd is to facilitate discourse cohesion by providing or reinforcing anaphoric co-
reference. Recalling the statement of Apollonius that second person pronouns are non-
anaphoric, we may interpret sd in these passages as rendering an inherently nonanaphoric
person explicitly anaphoric.

2. The foregoing discussion has centered on the role of anaphora as an exponent of textual
cohesion in the absence of overt conjunction. However, anaphora plays an important role
as a cohesive operator in constructions where overt conjunction is also found. In the
remainder of this paper we shall explore some of the interrelationships between anaphora
and conjunction in the Rigveda.

One of the most frequent tropes employed by the Rigvedic bards is relativization. Although
the relative pronoun in the Rigveda frequently possesses a specifying or individualizing
value (cf. Gonda 1954), relativization is often purely stylistic, as can be seen by a com-
parison of VIII. 63.6b Kkrtáni kártvàni ca ‘the things which have been done and are to be
done’ and 1.25.11c krtäni yd ca kartvä “id.’. Such relativization serves to facilitate the crea-
tion of the fundamental discourse strategies yd-... sd/td- and yd- ... noun together with
their inverted variants, as noted above. Beside the type X ya- ca Y we also find a syntagm
yá- X yá- ca Y. That this trope need not represent a true relativization of the conjoined
members can be seen from a passage such as I. 101.6a yah sürebhir hávyo yás ca bh irubhih
“Who is to be called by the heroes and by the cowardly’, where the conjoined terms are
surebhih and bhirubhih, neither of which is syntactically relativized. Rather, by an op-
tional stylistic transformation which we shall designate anaphorization, each of these terms
is made into an argument of a relative clause. Schematically, *yáh surebhih . . . bhirübhis
ca — ydh sürebhih ... yd$ ca bhirübhih. The result is stylistic anaphora together with ex-

197

plicit conjunction. Moreover, within the larger framework of the full stanza in which this
pada occurs we find anaphora qua coreference at a higher level of discourse structure
(yah ... yas ca/yah... yas ca // indram..../... havamahe, an expansion of the basic
coreferential type yd- ... noun). At the same time, pädas a and b constitute a perfect
vertical isosyllabic responsio (for the terms, cf. Gonda 1959: 166—176), itself a type of
anaphora involving complete syntactic constructions:?

L 101.6 yáh sürebhir hávyo yds ca bhirübhir | yó dhávadbhir hüyáte yás ca jigyübhih //
indram .../maritvantam sakhyaya havamahe ‘Which one is to be invoked by heroes and
by the cowardly, who is called by the fleeing and by the victorious, Indra . . . do we call for
friendship together with the Maruts.’

The process of anaphorization just described is not limited to stylistic relativization. In
particular, it includes the anaphoric use of conjunctions themselves. Thus, we find such
Rigvedic syntagms as X ca Y ca (frequent) and utá X utá Y (infrequent), e.g. I.25.20b
divás ca gmás ca ‘of heaven and of earth’ and VII. 98.2c uta hrdôtà mänasa ‘with heart and
with mind’. The issue concerning the diachronic relationship between the type with single
and double conjunction (cf. Dunkel 1982 with literature on ca) is thus a pseudo-problem.
The latter is derived from the former by the universal principle of stylistic anaphora. We
are not speaking here so much of a real diachronic priority as of a simple implication: given
X Y ca and X utd Y, the possibility of both X ca Y ca and utd X utd Y is immediate.”

In the passages discussed so far, anaphora works in concert with overt conjunction in order
to reinforce the conjoined structures (ya- X ya-ca Y, X ca Y ca). In many other instances,
however, anaphora and overt conjunction operate at different levels of structure, with
anaphora usually serving as a higher level discourse operator. This interplay of the two
processes is particularly clear at VII.92.3cd ni no rayim subhójasam yuvasva / ní virám
gdvyam asvyam ca radhah ‘Grant to us wealth bringing good enjoyment, a male (offspring),
(and) success consisting of cows and horses’. Here we find a two-tiered cohesive structure.
The anaphora ni... nf works in harmony with the metrical structure to demarcate the
objects of ni yuvasva: rayim in pàda c and the asyndetic virám . . . rádhah in d. The overt
conjunction, on the other hand, conjoins only the adjectival modifiers of radhah (gdvyam
dsvyam ca), which themselves illustrate morphological anaphora (both are denominal ad-
jectives in -ya) and homoioteleuton. A three-tiered anaphoric structure is seen in the fol-
lowing passage:

1.112.23 yabhih kutsam ärjuneydm satakratü / prä turvitim pra ca dabhitim ävatam A
yabhir dhvasäntim purusäntim ävatam | tabhir ü su ütibhir asvind gatam "With which (aids)
you two aided Kutsa, the son of Arjuna, you possessors of a hundred determinations,
Turviti and Dabhiti, with which you aided.Dhvasanti (and) Purusanti, with those (very)
aids, O Asvins, come hither.’

Within this passage the tightest cohesions, reinforced by both rhyme and meter, are prd
turvitim prä ca dabhitim and dhvasantim purusäntim, the former accompanied by stylistic
anaphora plus ca, the latter asyndetic. At a higher level, the outer anaphora yabhih ...
avatam //yabhih ... avatam defines the enclosed structures as hierarchically equivalent.
Here the anaphora is not merely lexical, but also syntactic: a constructional pattern is
repeated. The entire stanza is then defined as a unit by the coreferentially anaphoric struc-
ture yabhih .. . tabhir ü su.

198

A pattern showing both overt conjunction and lexical anaphora occurring at different levels
of structure is seen in the following instance:

11.13.7 ydh puspinis ca prasvàs ca dharmanà / ddhi dine vy àvánir ddharayah // yds casama
djano didytito divah |... sasy ukthyah ‘Who by reason of duty hast distributed the blos-
soming and fruitbearing (plants) (and) the streams over the meadow, and who hast begotten
the incomparable lightnings of heaven . . . as such, thou art praiseworthy.’

In this passage the overtly conjunctive anaphorization puspinis ca prasväs ca forms a tightly
cohesive unit on both the syntactic and semantic level. To this is then appended the asyn-
detic avánih of pada b. This entire conjoined structure, occurring only within the initial
distich (X ca Y ca/ Z), functions as a constituent within the larger conjoined structure
signalled by the outer anaphora (both lexical and referential) yah .. . yas ca, which informs
each distich. Typically, the entire structure is rendered cohesive at the outermost level of
the stanza by the referentially anaphoric sequence yéh... yah... sa.

Although the cohesive function of anaphora in both its senses has been illustrated with
evidence from the Rigveda, we are dealing with a universal principle of rhetorical style,
both poetic and prosaic. The interrelationships between the two types of anaphora are
particularly clear in the Rigveda, owing to the importance of the yd- . . . sá/tá- structure in
this text, with its possibilities for expansion (e.g. yd- ... yád- .. . sá/td-, yá- .. . sáftá- . ..
sa/td-, etc.), as well as to the frequent repetition of preverbs, negative particles, other ad-
verbials, and verbs. These figures are of great interest for the way in which they comple-
ment other ‘outils de structuration’ within the Rigveda, specifically the rich system of con-
junctions: (primarily) word-level ca, generally coordinating utd, originally anaphoric u,
clausal and interstanzaic ddha and dtha, etc. (cf. Klein, 1978b, 1980, 1985). To these
structural signals we may add the strong tendency for the major Rigvedic metrical units
(the pada, distich, and stanza) to coincide with important units of thought or discourse
(cf. Migron 1985). Taken together, these factors render the discourse structure of the
Rigveda, for all the complexities offered by the meaning of the text itself, in some sense
simpler and more transparent than that of other archaic Indo-European poetic texts.

Notes
: Dionysius seems not to have considered interrogatives, indefinites, and anaphorics to be true pro-
nouns, since when discussing the pronoun (one of his eight parts of speech) he refers only to per-
sonal, possessive, relative, and reflexive forms.

Although this treatise is traditionally ascribed to Demetrius of Phaleron, an Athenian peripatetic
philosopher and statesman of the late fourth century B.C., Roberts concludes, on the basis of both
internal and external evidence, that it is more likely to have been composed in either the first cen-
tury B.C. or the first century A.D., “the latter period being on the whole the more likely” (1902: 64).
Wackernagel (1928: 84) states that “Windisch hat das Verdienst in seiner Abhandlung über das Re-
lativpronomen . .. diese Unterscheidung [sc. anaphoric vs. deictic] samt dem Terminus ‘anaphorisch’
wieder ans Licht gezogen”.

For the relationship between the terms ‘annunciatory’ and ‘cataphoric’ cf. Klein, 1985 part 1: 17.

In her review of my book on u (Klein 1978a), Madame Bader (1979) rightly emphasizes the unitary
nature of anaphora in both its senses, hence the unitary function of u. However, she goes too far, in
my view, in rejecting the anteriority of the particle’s coreferential function relative to its conjunctive
value. The crucial point is that coreferential yd- ... sd/td- u cannot be interpreted as conjunctive
(*‘which one ... and that one’), whereas coreferential-iterative tdm ... fam u can be so viewed
(‘that one ... and that one’). See now Klein ms. for a parallel development in the diachronic syntax

199

of the cognate Homeric Greek particle ad. For another instance in the Rigveda where surface-level
ambiguity serves as the catalyst for syntactic change, see Klein 1982.

It is important to note that anaphora in the larger sense is not limited to lexical repetition, but may
include iteration at any linguistic level. Thus, morphological iteration and its frequently associated
rhyme or (when it involves contiguous terms) homoioteleuton (e.g. VIL 63.1c mitrásya várunasya)
may be considered a subtype of anaphora in which the repeated stretch is limited to only a part of
a lexical item. The same may be said of syntactic repetition or structural parallelism, where the
repeated matter involves an entire word group. I. 101.6ab provides examples of anaphora occurring
simultaneously at all these levels.

Since the type X ca Y ca is reconstructible for Proto-Into-European, it probably provides support for
the rare utd X utá Y (10 times, including the polysyndetic utá X1 utá X5... utá Xy). The latter syn-
tagm illustrates the appropriate position for a pre-positional conjunction, the former, for a post-
positional conjunction.

References

Bader, Francoise, 1979. “Review of: The particle u in the Rigveda. A synchronic and diachronic study by
Jared S. Klein," BSL 74 2: 128—131.

Bloomfield, Leonard, 1933. Language (New York: Holt, Rinehart and Winston).

Burchfield, R.W. (ed.), 1972. A supplement to the Oxford English Dictionary (Oxford: The Clarendon
Press).

Dunkel, George, 1982. “The original syntax of conjunctive *-KWe,” Die Sprache 28: 129-143.

Gonda, Jan, 1954. “The original character of the Indo-European relative pronoun io-,” Lingua 4:1—41.

Gonda, Jan, 1959. Stylistic repetition in the Rigveda (Verhandelingen der Koninklijke Nederlandse Aka-
demie van Wetenschappen, Afdeling Letterkunde. Nieuwe Reeks, Deel 65, No. 3). (Amsterdam: N.V.
Noord Hollandsche Uitgevers Maatschappij).

Halliday, M.A.K./Angus McIntosh/Peter Strevens, 1964. The linguistic sciences and language teaching
(Bloomington: Indiana University Press).

Jesperson, Otto, 1914. A modern English grammar on historical principles. Part 2: syntax, Vol. 1. (Hei-
delberg: Winter).

Klein, Jared S., 1978a. The particle u in the Rigveda. A synchronic and diachronic study (= Ergänzungs-
hefte zur Zeitschrift für vergleichende Sprachforschung, Nr. 27) (Góttingen: Vandenhoeck & Rup-
recht).

Klein Jared S., 1978b. “The system of coordinate conjunctions in the Rigveda”, IIJ 20: 1—23.

Klein, Jared S., 1980. ““Athä, ddha, and a typology of Rigvedic conjunction", 7/J 22: 195—219.

Kiein, Jared S., 1982. “Rigvedic tú and sú”, Die Sprache 28: 1—26.

Klein, Jared S., 1985. Toward a discourse grammar of the Rigveda. Volume 1: coordinate conjunction.
Parts 1 and-2. (Heidelberg: Winter).

Klein, Jared S., ms. “Homeric Greek ai: A synchronic, diachronic, and comparative study”, to appear
in the Zeitschrift fiir vergleichende Sprachforschung.

Liddell, Henry George/Robert Scott, 1968. A Greek-English lexicon (Revised and augmented through-
out by Sir Henry Stuart Jones with the assistance of Roderick McKenzie) (Oxford: The Clarendon
Press).

Migron, Shaul, 1985. The Rgvedic stanza as a syntactical unit: a study of selected trimeter passages.
Unpublished doctoral thesis, The Hebrew University.

Murray, James A.H./Henry Bradley/W.A. Craigie/C.T. Onions, (eds.), 1933. The Oxford English dic-
tionary (Oxford: The Clarendon Press).

Roberts, W. Rhys, 1902. Demetrius on style (Cambridge: University Press).

Schneider, Richard (ed.), 1878. Apollonii Dyscoli quae supersunt Vol. 1, Fasc. 1 (Leipzig: B.G. Teubner).

Uhlig, Gustav (ed.), 1883. Dionysii Thracis ars grammatica (Leipzig: B.G. Teubner).

Wackernagel, Jacob, 1928. Vorlesungen über Syntax. Zweite Reihe, zweite Auflage (Basel: Verlag Emil
Birkháuser & Cie).

Windisch, Ernst, 1869. Untersuchungen über den Ursprung des Relativpronomens in den indogermani-
schen Sprachen (Leipzig: C.P. Melzer).

The importance of Saussure’s Mémoire” in the development
of historical linguistics

Konrad Koerner (University of Ottawa)

0.0 Introductory Observations

Recent translations into Russian (1977) and Italian (1978) of Ferdinand de Saussure’s
Mémoire sur le système primitif des voyelles dans les langues indo-européennes, which had
first appeared in Leipzig in December 1878 (Saussure 1879), may be taken as an indication
of the continuing interest in and appreciation of Saussure’s contribution to comparative-
historical Indo-European linguistics.* I do not know of any other book in linguistics of the
last quarter of the 19th century that has been translated in recent years, with the exception
of Hermann Paul's Prinzipien der Sprachgeschichte of 1880 (5th revised edition, 1920),
which had translations into Russian and Japanese in 1960 and 1965, respectively. But
Paul's book was a contribution to general linguistics rather than historical linguistics as the
title of his book may suggest, and as such it was much less quickly outdated as is common
in a field in which new discoveries and advances make yesterday's findings obsolete tomor-
row. The fact that Saussure's Mémoire did not share the fate of most other studies of the
period calls for special reasons, and I am hoping that the present paper will offer at least a
few of these.

0.1. Early reactions to Saussure’s ‘Mémoire’

There are indications that Saussure’s 300-page study on the sound system of Proto-Indo-
European made a notable impact on linguistic circles of the time, though perhaps not the
impact that the young author might have hoped for. We may recall that when the book was
published, Saussure was just 21 years old, and a student of linguistics at the University of
Leipzig for barely four semesters; indeed, when he defended his thesis on the absolute
genitive in Sanskrit in February 1880, it is said that one of his examiners assumed that he
was a young relative of the well-known author of the Mémoire! But apart from Saussure’s
precocity, there were other factors that hampered the immediate success and the accept-
ance of the ideas he proposed in the Mémoire. Despite the fact that he was a student of the
Junggrammatiker, notably Karl Brugmann (1849—1919), the prime mover of this group of

* Revised version of a paper first presented at the Sixth International Conference on Historical Lin-
guistics held in Poznan, Poland, on 22—26 August 1983. It is with distinct pleasure that I dedicate it
to Henry M. Hoenigswald as a token of my admiration for his scholarship as an historical linguist as
well as an historian of linguistics.

202

young revolutionaries in historical-comparative linguistics at the time, Brugmann did not
concur with the main tenets of Saussure’s argument, although he was impressed by his
breadth of coverage, circumspection, and “nicht gewöhnliche Combinationsgabe”. Indeed,
Brugmann (1879a: 774) felt that Saussure had proposed a purely aprioristic scheme (“rein
aprioristische Construction”), which did not hold water, and he was not convinced that
Saussure’s proposals would have to lead him to a substantial revision of his own views (cf.
also Brugmann (1879b: 261—262), and the relevant quotations in Redard (1978: 33)).

Hermann Osthoff (1847-1909), Brugmann’s close collaborator during the 1870s and
1880s, expressed himself in a much more hostile manner to Saussure’s theories in several
articles published in volumes 2 and 4 of Morphologische Untersuchungen in 1879 and
1881, qualifying them as a ‘total failure’, ‘radical error’, and the like (cf. Redard (1978:
35) for details).' Gustav Meyer (1850—1900), another member of the group, incorporated
part of Saussure's findings in his Griechische Grammatik of 1880 (3rd ed. 1896), without
ever mentioning the source of the series of the Indo-European ablaut he was basing his
work on. In short, it can be said that, at least in its initial phase, Saussure’s Mémoire did
not find the reception within the immediate circle of the Leipzig linguists one might
perhaps have expected. However, other reviewers of the book who were either opposed to
the Young Turks at Leipzig or abroad and thus far removed from the scientific quarrels
among German scholars, seem to have agreed on at least one trait of Saussure’s Mémoire,
namely, that it was theoretically very demanding, that the author had a penchant for alge-
braic formulae, and that the argument was difficult to follow because of a predilection for
abstraction. Thus August Fick (1833—1916), the head of the Góttingen linguistic circle
which included, inter alia, Hermann Collitz (1855—1935), reviewing Saussure's book at
considerable length, noted Saussure's fondness of “mathematische Formulierungen”
(1880: 420), but concluded, after having criticized a number of points of detail in Saus-
sure's argument:

“Die etwas künstliche Anordnung des Stoffs und eine eigentümliche Vorliebe des Verf[as-
sers] für mathematische Formulierungen erschweren das Studium des sonst klar geschriebe-
nen Werkes, wer aber die Mühe nicht scheut, wird sich durch mannigfache Belehrung und
Anregung reichlich entschádigt finden." (Fick 1880: 239)

The Celtologist, (Sir) John Rhys (1840—1915), writing for a more general audience in a
British monthly, likewise noted, in his very positive review:

"The work is so technical and the reasoning so complicated that it is very difficult to give
any idea of it excepting to those who have had a thorough training in Aryan glottology
[i.e., Indo-European linguistics]; but it may at once be said that it is the most important
and epoch-making work that has appeared since [Johannes] Schmidt's [(1843—1901)] [Zur
Geschichte des indogermanischen] Vocalismus |2 vols., Weimar: H. Böhlau, 1871—75] was
published.” (Rhys 1879: 234)

We may also cite a similar opinion by the French Latinist and comparative linguist Louis
Havet (1849-1925), who in fact wrote the most extensive review of Saussure's Mémoire,
but had similar misgivings, arguing that it had “un defaut grave: il est extraordinairement
dur à lire”. Havet found that Saussure was abusing abstract designations and was engaging
in an excessive use of (at times newly coined) technical terms, concluding his complaints by
stating that “Ainsi, M. de Saussure fait suer sang et eau à ceux qui le lisent” (Havet 1978
[1879]: 107). However, apart from these criticisms, Havet’s review of Saussure’s book was
very favourable indeed, leading to a life-long friendship and correspondence between the
two scholars (cf. Redard 1976).

203

There is still at least another review of the Mémoire that deserves being mentioned, namely,
the one by the young Polish linguist Mikołaj Kruszewski (1851—87), which appeared in
Russian in 1880. It also included an account of Brugmann’s “Nasalis sonans” article of
1876, but it is obvious that Saussure’s Mémoire was far more attractive to the theoretically
inclined Kruszewski, who regarded Saussure’s book as an important contribution to the
method of phonology, and in fact as marking a new phase in the study of Indo-European
phonology (Kruszewski 1978 [1880]: 444 and 450, respectively). (1 may also mention in
parenthesis that the Mémoire inspired Kruszewski’s work in various ways; apart from
leading him to study questions of ablaut in Indo-European, especially in Slavic, Kruszewski
— and his mentor and collaborator Baudouin de Courtenay — took from the Mémoire a
number of concepts and terms, such as ‘phoneme’, ‘alternation’, and ‘zero’. Indeed, Kru-
szewski translated ‘phoneme’ into Russian as ‘fonema’ and proposed to use it as a pho-
nological term in contradistinction to ‘zvuk’ “sound”, a distinction which made history
in linguistics.)

In sum, we may say that outside the narrow circle of the Leipzigers Saussure’s work was
well received when it was first published. It is true that in France Saussure’s proposals
found an untiring critic in Paul Regnaud (1838—1910), a professor of Classics and com-
parative grammar at the University of Lyon (e.g., Regnaud 1889, 1890, 1891), but it seems
that Regnaud’s attacks on Saussure’s system of the proto-Indo-European vowels were
ignored by most of his contemporaries, probably for the simple reason that the younger
generation of Indo-Europeanists had long since rejected the basic assumptions of August
Schleicher (1821—1868) in matters of Proto-Indo-European phonology, to which Regnaud
still subscribed.

0.2. The reception of Saussure’s ‘Mémoire’ during the late 19th and early 20th century

Ignoring for the moment the peculiar reception Saussure’s theories received from scholars
such as Hermann Möller (1850-1923) and Albert Cuny (1869-1947) — it will concern us
later in this paper — we may note that members of the next generation of Indo-European-
ists, notably Saussure’s former student at the Ecole Pratique des Hautes Etudes in Paris,
Antoine Meillet (1866—1936), and two German scholars, Wilhelm Streitberg (1864-1925)
and Hermann Hirt (1865—1936), had a particularly high opinion of the Mémoire. Hirt can
be said to have been the first scholar to fully incorporate Saussure’s findings in his mono-
graph of 1900; already in 1885, the brilliant comparative linguist Heinrich Hübschmann
(1848—1908), one of Saussure’s former teachers at Leipzig, was close to accepting Saus-
sure’s theory part and parcel had it not been for an error on Saussure’s part in taking Lat.
ago and the like as an aorist present of *Ag- (Hiibschmann 1885: 2—3, cf. Mayrhofer 1981:
26-27, n. 75). Hirt, for his part, regarded Saussure’s Mémoire as “ein bahnbrechendes
Werk, noch heute von größter Bedeutung”, some two generations after its publication (Hirt
1927: XXV), thus echoing Streitberg’s opinion expressed in his obituary of Saussure of
1914:

“Das Buch ist de Saussures Meisterwerk: noch heute, nach einem Menschenalter, wirken In-
halt und Form mit derselben bezwingenden Macht wie am Tage des Erscheinens — von wie-
viel sprachwissenschaftlichen Werken, auch solchen höchsten Ranges, kann man das Glei-
che sagen?” (Streitberg 1966 [1914]: 103)

204

Last but not least, reference should be made to Meillet, in particular to his influential
Introduction à l'étude comparative des langues indo-européennes, which first appeared in
1903, and which bears the inscription “A mon maître Ferdinand de Saussure à l’occasion
des vingt-cinq ans depuis la publication du Mémoire . . .(1878—1903)”, an inscription that
was retained in the many subsequent editions of the text. As a matter of fact, this book
remained for Meillet Saussure’s major achievement, and not the posthumous Cours de lin-
guistique générale, with the result that even after the latter’s appearance, Meillet associated
the well-known phrase (wrongly ascribed to Saussure, though undoubtedly in his spirit)
that language is a system ‘où tout se tient’ with the Mémoire. (Cf. Meillet (1903; gth edi-
tion (1937), p. IX [= 1903: X], 475), and elsewhere; and Brogyanyi 1983, for details.)

In 1887, Saussure's Mémoire was republished in Paris; at this time, Saussure contemplated
accompanying the new edition with an attack on his detractors (cf. Redard 1978: 36). One
may speculate who or what circumstances persuaded Saussure to suppress his ill feelings
toward those who did neither understand nor appreciate his accomplishments. It could
have been the recognition he received from scholars outside the narrow circle of the Jung-
grammatiker, notably Havet, Fick, and Hübschmann, all between 9 and 24 years his seniors,
and perhaps by the opinion expressed by the doyen of Indo-European linguistics at Leipzig,
Georg Curtius (1820—85), in a personal letter of 1884: “Lange habe ich nichts der Art ge-
lesen, was mich so entschieden überzeugt hat."?

0.3. The third phase in the reception of the ‘Mémoire’

The phase following Saussure’s death in 1913 may well be regarded as the most important
period in the development of the discussion of the Proto-Indo-European phonological
system, a phase which has not yet come to a close. Saussure’s Mémoire may be said to have
played an important role in this discussion, which in fact began as early as in 1879, though
more as a fairly weak undercurrent as we may see later on. There were a number of reasons
for the slow posthumous recognition of Saussure’s work in Indo-European linguistics. For
example the Cours of 1916, owing to its seeming anti-historical bias, tended to detract
from the fact that Saussure’s life-time contribution lay almost exclusively within dia-
chronic linguistics. Also the hostile feelings toward anything German of the period during
and following the First World War made the work of a French Swiss scholar rather at-
tractive as it appeared to oppose the traditional kind of linguistics widely associated with
Germany. As a result, among many of the new generation of linguists historical linguistics
became a side-issue and the preoccupation of a comparatively small number of experts.
This development is clearly reflected in most histories of linguistics. If they mention Saus-
sure’s Mémoire at all, they present it as something of marginal interest, focussing instead
almost their entire attention on the work that in fact did not have his imprimatur and, as
we have come to understand in recent years, does not reflect his true intentions. In short,
it has become customary to refer to a Saussure who wrote the Mémoire in his youth, and
another who is the author of the Cours, which brought about a revolution in modern lin-
guistics, with the result that even the large portions devoted to diachronic linguistics it
contains are patently ignored.

It would exceed the frame of the present paper to show, as Cristina Vallini (1969: 49ff)
has already done to some extent, that in fact there is no justification for the traditional

205

conception of the existence of ‘deux Saussure’ (cf. Redard 1978), but that the early work
is consistent with his later, largely unpublished theories. Perhaps it should be stated in the
present context that the critical edition of the Cours, carefully compiled by Rudolf Engler,
contradicts affirmations in the text as edited by Bally and Sechehaye, including those fre-
quently attacked ones according to which synchrony and diachrony are supposed to be
regarded as two subjects apart: Nothing could have been more contrary to Saussure’s view
on the proper relationship between these two ‘points de vu’.

We know from the Cours the central role that the concept of ‘system’ has played in Saus-
sure’s linguistic theory. But if we analyse his Mémoire carefully, we will notice that the
same notion is basic to Saussure’s argument therein as well. Indeed, we may say that in the
Mémoire ‘system’ is the ‘clé de votite’, the key stone holding the entire edifice together.
Interestingly enough, while others referred to Saussure’s book as the Mémoire for short, he
himself preferred to refer to it as his Systéme des voyelles, with a capital ‘S’ (cf. Watkins
1978:61). Moreover, Saussure justified the title of his book in the following terms, at the
same time indicating his approach to the subject matter:

“Étudier les formes multiples sous lesquelles se manifeste ce qu’on appelle l'a indo-euro-
péen, tel est l’objet immédiat de cet opuscule:le reste des voyelles ne sera pris en considéra-
tion qu’autant que les phénomènes relatifs à l’a en fourniront l’occasion. Mais si, arrivé au
bout du champ ainsi circonscrit, le tableau du vocalisme indo-européen s'est modifié peu à
peu sous nos yeux et que nous le voyions se grouper tout entier autour de l’a, prendre vis-à-
vis de lui une attitude nouvelle, il est clair qu’en fait c’est le système des voyelles dans son
ensemble qui sera entré dans le rayon de notre observation et dont le nom doit être inscrit
à la première page.” (Saussure 1879[1878]: 1; emphasis added: KK)

Similarly, and quite revealingly I believe, Saussure ends his book with the following para-
graph, in which he discusses special instances of the possible presence of this elusive Proto-
Indo-European a or, rather, in Saussure's analysis, e:

"Cette inconstance de la voyelle [in western Indo-European languages] révélerait, dans
d'autres circonstances, la présence du phonàme 4; mais si telle est la valeur de l’e dans
&FéEc, la relation de cette forme avec vákSati, ukYáti, aU£co, aussi bien que sa structure
considérée en elle-méme cessent d'étre compréhensible pour nous." (Saussure 1879: 283;
emphasis added: KK)?

This concept of system is not simply what is implied by expressions such as ‘le système de
Curtius’ (Saussure 1879: 2), 'systéme de Schleicher' (1879: 3—4), and the like, which refers
merely to the hypothesized inventory of the sounds of a language, but, as will become
evident in the subsequent discussion, a system of relationships of phonemes. Indeed, we
may discover in Saussure's Mémoire the source of his later definition of phonemes as “des
entités oppositives, relatives et négatives" (Saussure 1916: 164; cf. Saussure 1968: 268).
Already the first paragraph of his Mémoire cited above suggests this.

1.0. Saussure's argument in the ‘Mémoire sur le système primitif des voyelles dans les
langues indo-europeennes’

Before delineating the manner in which the Mémoire made an important contribution to
linguistics, both historical and theoretical, let us first present the main tenets of Saussure’s
argument, his approach having been signalled in the introductory paragraph of his work,
namely, to investigate the various reflexes in the attested languages of what has been com-

206

mon Indo-European a, with all other vowels of the system being considered only in relation
to this phoneme, but with the result that eventually the entire system of Indo-European
vowels is established.

Saussure proceeds in a most rigorous fashion. He discusses — and discards — one proposal
after another made by Curtius, Schleicher, Fick, and others, in part also those by his
teachers, including Brugmann and Osthoff. Subsequently, he adduces evidence for his own
views, and this with a precision that must have astounded his contemporaries, and which
impresses us still today. I am quoting just one example from the English translation, pro-
vided by Winfred P. Lehmann in 1967, to illustrate Saussure’s style of scientific discourse.
In this passage from chapter 2 of his Mémoire in which he proposes to separate the ‘old’
from the ‘true’, by lifting “tout l’humus moderne que différents accidents avaient amassés
sur lui”, Saussure, after having established the various correspondences between root
vowels found in what he terms the northern languages, e.g., Germanic and Slavic, and the
southern languages, such as Latin and Greek, proceeds as follows:

“How then could the a and o of the languages of the South have arisen from one and the
same original a? By what miracle could this old a be coloured to o, and never to a, precisely
in all the times that it is found in the company of an e? — Conclusion: the dualism of a and
o is original, and it must be that in the single a of the North two phonemes were confused.

Confirmation: when a root contains a in Greek or Latin, and when this root is found in the
languages of the North, one observes in the first place that this a never alternates with e, as
is the case when the Greek replies by an o. Thus Goth. vagja = Gk. okhéo, hlaf =Gk.
(ké)klopha are accompanied by viga and hlifa. But agis(a-) = Gk. dkhos, or ala = Lat. alo do
not have a parent form with e. On the other hand, the roots of the latter type have one
characteristic, unknown to the first type: the ability to lengthen their a (agis: og, ala: ol),
of which we wiil have to take account below.

Brugmann has designated with 4, the prototype of the European e; his 4; is the phoneme
which we have called o up to now. As to this third phoneme which is the Greco-Italic a and
which constitutes one half of the a's of the languages of the North, we will designate it by
the letter A, after noting well that it is neither the e(2,) nor the o(a;). — Excluding for the
time being the other possible kinds of a, one obtains the following table:*

Langs. of the North Primordial state Greco-Italic

At first sight, it may appear to be an over-interpretation to claim that Saussure’s use of
‘state’ (“état”) in this table has any significance. But if one follows his argument more
closely, it becomes evident that he is primarily concerned with establishing an ‘état de
langue' in his Mémoire, namely that of the Proto-Indo-European vocalic system, not its
succession. Indeed, one indication of this intention is Saussure's regular use of the term
*phonéme' in his work instead of ‘son’, a usage which is the first of its kind in the history
of linguistics, even though the term itself had not been invented by Saussure, as Hjelmslev
(1970: 125) holds. Saussure wanted to signal by this that he is not concerned with matters
of phonetic realization; as a matter of fact, he rarely makes a reference to phonetic detail
in his book (cf. Polomé 1965: 10, and n.7), preferring instead an algebraic notation devoid
of phonic substance. From what will become clear below, we may say that Saussure’s

207

Mémoire is structuralist in approach, and that it constitutes in essence a contribution to
structural linguistics ‘avant la lettre’ and not merely to historical-comparative linguistics as
it is commonly understood.

However, this understanding of the Mémoire is of a much later date, and it required the
work of kindred spirits such as Louis Hjelmslev (1899—1965) to fully appreciate this
endeavour on Saussure’s part (cf. Hjelmslev 1970: 123—128, for details). In the context of
the present paper this essential trait of Saussure’s argument must remain a matter of lesser
consideration, although it should be clear that his theory of Proto-Into-European ablaut
cannot be understood without the concept of language as a rigorously organized system
underlying it (cf. also Schmitt-Brandt 1967: 2; Vincenzi 1976). For the field of Historical
Linguistics, which at the time was almost identical with linguistics tout court, Saussure’s
Mémoire must now be regarded as introducing a revolution, even though it is safe to speak
of ‘une victoire retardée’ or ‘tardive’, since, as we have seen in the introduction (0.1.—0.2.
above), most of Saussure’s contemporaries did not grasp the full implications of his theo-
ries, and a number of distinguished scholars, including Brugmann, did not see the necessity
of revising their inventory of the Proto-Indo-European sounds in accordance with Saus-
sure’s findings. As will become clear from what follows, however, there were, apart from
Saussure’s quasi-algebraic approach to the subject, other reasons which explain the dif-
ficulty Saussure’s contemporaries, not least the Junggrammatiker with whom he had been
associated for some time, experienced in accepting his proposals.

1.1. Saussure’s revisions of the Proto-Indo-European vocalic system

We have already hinted (1.0. above) at Saussure’s general procedure in dealing with the
problem of the Proto-Indo-European vocalic system. However, we cannot fully appreciate
his in fact audacious innovations unless we present ourselves what had been, until the
mid-1870s, received opinion.

Until about that period, Indo-European linguistics, despite a number of statements by their
proponents to the contrary, was essentially ‘Sanskrito-centric’ (cf. Mayrhofer 1983: 130—
136 passim). This meant that, in matters of phonology especially, linguists clung to the
belief that Sanskrit, which, until the decipherment of Hittite during the early decades of
the 20th century, supplied the comparative linguist with the oldest data, was in effect
reflecting the earliest state of Indo-European. As a result, the inventory of Proto-Indo-
European vowels was regarded as consisting of three vowels only; which could either be
long or short:
i u
a

Since Greek and Latin for instance displayed a five-vowel system, this was accordingly inter-
preted as representing an innovation on the part of these (classical) European languages,
usually regarded as a degeneration of the primordial state; cf. the following correspondences
taken from Arlotto (1972: 118):

Sanskrit Greek Latin Indo-European
(a) ad-mi ‘I eat’ (e) ed-omai ‘I will eat? — (e) ed-o leat (*e) *ed- *to eat'
(a) pra-ta ‘full’ (é) plé-rés ‘full’ (e) ple-nus ‘full’ (*e) *ple- ‘to fill’

(a) asta ‘eight’ (0) okto ‘eight’ (0) octo ‘eight’ (*o) *okto ‘eight’

208

and so on, where there are still two other correspondences in Greek and Latin to Sanskrit
a and 4, which are also a and @, respectively. The i/u/a vowel triad, however, had been
codified in Schleicher (1861: 134—135), and was widely accepted for several years after
Schleicher’s death in 1868. Indeed, his work was reedited twice (1871 and 1876) by two
of his former students, Johannes Schmidt (1843—1901) and August Leskien (1840—1916),
and thus continued its influence in the teaching of Indo-European phonology. The rejec-
tion of the Proto-Indo-European vowel triad and the acceptance of a much more variegated
system of both short and long vowels in Indo-European was one of the main accomplish-
ments of the younger generation of comparative-historical linguists during the later 1870s.
But Saussure was not satisfied with a simple enlargement of the vowel triad to a five-vowel
system as the classical languages may have suggested, namely:

However, this view of the Proto-Indo-European vowel inventory had become widely ac-
cepted about the same time that the Mémoire appeared. Saussure did not appear to sub-
scribe to this configuration of the vocalic system of Indo-European and, as a result, made it
difficult for his contemporaries to appreciate and, furthermore, to accept his own pro-
posals. As a matter of fact, these came, as Szemerényi (1980: 115) notes, at the worst
possible time.

Without adducing much evidence for his hypothesis, Saussure claimed (1879: 135) the
following: First of all, the basic root vowel in Proto-Indo-European is /e/ rather than /a/ or
a variety of vowels; this was the a; that we noted earlier, and which both in the northern
dialects and Greco-Italic are represented by e (cf. Saussure 1879: 52). He had earlier noted
that the high vowels i and u, like the nasals and liquids, could take on (semi-)vocalic func-
tion, and therefore (1879: 8) he put them together in one group as “coefficients sonanti-
ques": Depending on the presence or absence of the e within the root, r / m n become
sonants, and į and u take on consonantal function.” (We are long since used to calling these
sound ‘resonants’.) In other words, Saussure argued that all root vowels consist of combina-
tions between the basic Proto-Indo-European e and a resonant or ‘sonant coefficient’.
Under certain conditions, this e was ejected (quantitative ablaut), leading the resonants to
take on syllabic function, and under others, it alternated with o (qualitative ablaut),
represented by a; in earlier theories (such as those by Brugmann).

In addition to alternation with o, e could simply disappear altogether, thus producing the
following results at the “degré zéro” (‘zero grade”); example:
*peik-> *pik- *penk-> *pnk- and *pet- > *pt- (without a vowel)

In addition to the six sonorants or resonants mentioned above, however, Saussure hypoth-
esized the presence of two more “coefficients sonantiques”, which he symbolized as À and
Q respectively, both sounds which could appear on their own only at the so-called zero
grade of a root form, and would produce a and o respectively — the latter being of course
different from the o, in the ablaut relation (and should therefore be identified as 02) or —
and this should be highlighted here: shwa. Saussure (1879: 178) described the latter as a
“voyelle indéterminée” and as “une espèce d'e muet, provenant de l’altération des pho-
némes A et Q” (italics in the original).

209

These two added sonant coefficients play a central role in Saussure's argument, as he sets
up the following combinations to obtain the series of long vowels:

e+A=éanda, ande+0=0
As Saussure does not specify the phonetic quality of his 4, this ‘sonant coefficient’ can
produce both è and à; o can also be the result of the apophonic o + A oro + Q. At any rate,
all long vowels are products of the original e plus a coefficient, and hence not original as
having independent existence. Examples:

*bheA- = *bha- cf. Gk pha-mi (Attic Gk: phemi) ‘I say’

*bhA- (zero grade) cf. Gk phá-mén ‘we say’

*deQ- cf. Gk do- (root of) ‘to give’

*dQ- (zero grade) cf. Gk do-tos ‘given’
No doubt, Saussure operates with what we nowadays refer to as ‘underlying forms’, deriving
the actual attested forms through specific rules. By the same method, Saussure (1879: 248)
sets up the rule “Le groupe sonante + A, précédé ou placé au commencement du mot, se
change en sonante longue, quel que soit le phonème qui suit” (italics in the original), so
that / and 4 as well as the long sonorants à in Î î are derived from iA, uA, nA, rA, and so
on, or, in notation suggested by Saussure only in 1891 (cf. Saussure 1922: 603), sonant
plus shwa. In effect, it would seem that Saussure was the first abstract phonologist in that
he was operating with hypothetical constructs and indirect (distributional) evidence.

From his Mémoire at least, it is quite clear that the shwa was supposed to have vocalic
rather than consonantal quality, and Mayrhofer (1983: 144n.91) may be justified in his
hesitation to accept the view expressed by others that in his 1891 observations Saussure
was in fact modifying his earlier interpretation. This question is relevant in the subsequent
discussion of Proto-Indo-European phonology (cf. 1.2. below); but what is important to
emphasize in the present context is that Saussure had developed a fairly abstract and rigid
system of phoneme combinations which allowed him to produce the various attested forms
in individual Indo-European languages. As Szemerényi (1980: 115) has pointed out, Saus-
sure did not present much in terms of evidence for his theory except for hinting at such in-
cidents in Greek and Sanskrit which may be explained by his assumptions. The beauty of
the matter is that much of what Saussure advanced on theoretical and structural grounds
was subsequently shown to be correct, to the extent that we may be right to say that with
Saussure’s Mémoire a new phase in the history of Indo-European phonology had been set
in motion.

1.2. Saussure’s contribution to the laryngeal theory

It is not surprising that Saussure’s discussion of Proto-Indo-European vowels contained one
or the other error, though none substantial enough to lead to a rejection of his theory as a
whole. We have already hinted at the fact that Saussure was not much concerned with
phonetic detail in his argument; this may have indeed been a point of strength rather than
weakness, as some modern phonologists may well be ready to admit. But in Saussure’s
time, when linguists were beginning to take note of articulatory phonetics, this attitude
might easily have been subject to criticism. August Fick, in his review of the Mémoire, was
one of the first to point to the difficulty of deriving both 4 and e from e*t A; I guess we
would today call Saussure’s hypothesis ‘counter-intuitive’ on this issue and lacking in sym-

210

metry. Saussure had considered such a combination of a,a,fee) “paralléle aux combi-
naisons 4,4, @ii, ain, [i.e., *eA, ei, en] etc.”, but he argued in fact that this would lead
to ‘contre-sens’ (1879: 141). Fick (1880: 437), however, felt, although he conceded that
his “Begeisterung fiir das Analogieprincip nur schwach ist” (1880: 436), that a ‘coefficient
sonantique’ £ parallel to Saussure’s A and O was indeed desirable, at least as long as one
admitted the original existence of an unaccentuated e in addition to Saussure’s ‘funda-
mental e”.

But as we noted in the introductory part of this paper (0.1.--0.3.), most Indo-Europeanists
of the last quarter of the 19th and the first quarter of the 20th century ignored Saussure’s
theory of sonant coefficients, and it was only some fifty years after the publication of the
Mémoire (and 14 years after his death) that the significance of Saussure's proposals was
fully recognized. The path of recognition was long and slow, and the fact that the first
scholars to see the light were outside the mainstream of Indo-European linguistics did not
help to further it.

Shortly after the appearance of Saussure's book, the Danish Nordist Hermann Möller
(1850—1923), who also worked in Semitic and in fact adhered to the Indo-European-
Semitic hypothesis so ardently combatted by Schleicher and dismissed by the generation of
Indo-Europeanists following him, proposed, in a review of a book by Friedrich Kluge
(1856—1926)

„ein consonantisches element ... , welches die eigenschaft hatte, ein vorangehendes oder
(im anlaut) folgendes 2, (das in der letzten zeit der grundsprache ein d gewesen sein wird)
in reines a zu verwandeln, und das mit vorhergehendem vocal a zu reinem langem 4 zusam-
menschmolz." (Móller 1879: 150).

In the same review, Möller suggested that in order to justify the alternation in Gk 0-/0e-,
it would be required that a third sound, E, should be added to Saussure's sonant coef-
ficients 4 and O (Móller 1879: 151), n.1), quite in line with Fick's suggestions mentioned
earlier. However, in contrast to Fick (1880: 438), for whom these coefficients were “von
Haus aus Vocale", Móller had hypothesized them to be consonantal in nature. Henry Sweet
(1845—1912), also reviewing Kluge's (1879) book, though together with Saussure's Mé-
moire, proposed by following Móller's lead to regard A as a voiced glottal trill, with E being
characterized by a palatalized, and O by a velarized articulation (Sweet 1880: 161). In the
meantime, Móller was shifting his position: While he had hypothesized “A [as] die tönende,
E die tonlose kehlkopfspirans?, O das kehlkopf-r?” in 1879 (Müller 1879: 151 n.1), he
connected the Indo-European sounds in question with Semitic in the following year by sug-
gesting that we have to do with “wahrscheinlich gutturale von der art des semitischen,
A = älef, der tonlose verschlußlaut, und E wahrscheinlich der entsprechende tönende ver-
schlußlaut” (Möller 1880: 492 n.2).

However, Möller did not draw the conclusions from his own observations for almost forty
years (cf. Möller 1917: 4-5 and 53), and since his 94-page monograph on the laryngeal
consonants of Indo-European and Semitic was not regarded as sound in scholarship with
regard to much of the data presented therein to bolster the claim for an early Indo-Euro-
pean-Hamito-Semitic genetic relationship, Möller’s proposals remained largely ignored. (For
a detailed analysis of Móller's suggestions, see Szemerényi 1973: 5—8.) Eventually, his
laryngeal hypothesis of 1879 found adherents, though not before the early years of the
20th century. Thus in 1907 his former student Holger Pedersen (1867—1953), at that time

211

sympathetic to the Semitic-Indo-European hypothesis, tried to adduce evidence that Proto-
Indo-European shwa” derived from a consonant By that time, Möller (1906) had pro-
posed, on the basis of Semitic, a theory of altogether five similar sounds, which, in the
preface to his Comparative Semitic-Indo-European Dictionary (Môller 1911: VI), he called
‘laryngeals’, in fact stating that “die von F. de Saussure fiir das Vorindogermanische er-
schlossenen ‘phonémes’ entsprechen den semitischen Laryngalen”. One may wonder
whether this remark and related observations by Möller led the French scholar Albert Cuny
(1869-1947), another strong believer in the original genetic relationship between Afro-
Asiatic and Indo-European (called ‘Nostratic’ by Pedersen), to assert that Saussure had, at
least since 1891, thought of the shwa as having consonantal function (Cuny 1912: 119; cf.
Szemerényi 1973: 8—10, for a thorough refutation of this common assertion).

But Cuny made a number of much more striking observations in his 1912 article, which
followed an earlier dissatisfaction with Móller's views on the relationship between Indo-
European and Semitic. Indeed, this article contains a series of important observations, of
which only few will be mentioned here (cf. Szemerényi 1973: 13—15 and Jonsson 1978:
3—4 for details):

(1) The Indo-European proto-language must have contained altogether three consonantal
laryngeals (E, A, O; a, a2, a3 Or Hi, Hy), H3, depending on notational preference) instead
of Saussure’s two vocalic Q and A. (The series of three such ‘coefficients’ had been sug-
gested as early as 1879 by Möller and 1880 by Fick.)

(2) Cuny allowed for a certain flexibility in the matter concerning the consonantal quality
of these laryngeals, however. Thus he had noted that in sequences of sonorant + laryngeal +
obstruent, it is always the sonorant that becomes vocalic, e.g., sfera- : stratos, which of
course means that the laryngeal is less inherently vocalic than the sonorant. However, in
the absence of a sonorant, the laryngeal, following the loss of a vowel, is vocalized, e.g.,
sag- standing for (in Saussure's notation) sedg- has its counterpart in sag- (for sag-), having
the formula sdg- underlying (Cuny 1912: 102—103). This flexibility allows Cuny (1912:
120) to show that Greek for instance had a series of three vocalic shwas.

(3) In addition to Saussure's suggestions, in particular the one made in 1891, according to
which an unvoiced spirant derives frequently from ¢ + laryngeal, Cuny added further evi-
dence for the original presence of laryngeals in Indo-European, in Indic as well as in Balto-
Slavic (Cuny 1912: 117—120; cf. Szemerényi 1973: 13—14).

Cuny's argument, if followed closely, reveals at least two major traits: first, he is very much
indebted to Saussure's (and also to Móller's) findings; second, we may note a considerable
increase in phonetic-phonological sophistication. What in addition deserves to be pointed
out is that mainly internal evidence, i.e., evidence deriving from attested language material
of Indo-European, was supplied for the laryngeal hypothesis. The fact that the laryngeal
theory, as it is more generally called today, had originally been inspired by findings made
in Semitic does not contradict this statement, though the fact that its main proponents
subscribed to the pre-Indo-European genetic relationship with Hamito-Semitic, appears to
have made it difficult for mainstream Indo-Europeanists to accept the theory. Therefore,
further evidence was required to lead the laryngeal theory to victory, and it came from a
former student of Antoine Meillet, Jerzy Kurytowicz (1895—1978), in 1927, two genera-
tions after the publication of Saussure’s Mémoire.

212

It is no exaggeration to state that Kurytowicz’s connection between Indo-European shwa
and a sound actually found in Hittite, an Indo-European language older than Greek or
Sanskrit as far as the available data is concerned, which could be compared with the
hypothesized Proto-Indo-European phoneme (Kurytowicz 1927), could not have been
undertaken without the work of another researcher. The evidence adduced by Kurytowicz
had not been available to either Saussure or his early followers like Móller, Cuny, or Peder-
sen. Indeed, it was only several years after Saussure’s death in 1913 that the tablets of
cuneiform inscriptions, discovered during excavations near the Turkish village of Bogaz
Köy during 1905-1907, had been deciphered and identified as Indo-European, namely,
by the Czech scholar Bedfich Hrozny (1879--1952) in 1915 and the following years.

Kurytowicz (1927) had examined Hittite words whose etymological connection with other
Indo-European languages such as Greek or Latin was beyond doubt. He chose examples
where Saussure had predicted the original presence of a sonant coefficient, in particular
the A, and could establish the following correspondences among others, where Hittite has
the laryngeal A:

Greek anti = Hittite kanti ‘in front? <*Aanti
Greek arges Hittite harkis ‘white’ <*Aarges
Latin pdsco = Hittite pahsanzi “protect <*paAsk-

The above examples prove another hypothesis made by Saussure, namely, that if (ori-
ginally) the laryngeal preceded the vowel, the result is a short 2, and a long one, if it
follows. As a result, it could no longer be doubted that early Indo-European contained
laryngeals at one time of its development, traces of which can be found in the various
attested languages. In this way, we may say that Kuryfowicz’s work vindicated Saussure’s
theory, which from then onwards has become more and more generally accepted by
historical-comparative Indo-Europeanists.

Of course, Kurytowicz’s paper of 1927 was not a one-shot enterprise; on the contrary, as
Szemerényi (1973: 15—19 and especially note 40) has shown, he published a flurry of
studies to establish his proof on a very broad basis of phonological evidence himself, and a
number of other scholars followed his lead in quick succession. Among them, mention
should at least be made here of Emile Benveniste (1902--1976) and his important work of
1935, in particular his theory of the basic structure of the Indo-European root, and Walter
Couvreur’s massive monograph of 1937, which for the first time accounted for the fact
that the Hittite laryngeal occured sometimes in a simple, sometimes in a double consonant.
Both Benveniste and Couvreur (cf. also the latter’s 1939 paper) based their argument
largely on Saussure's Mémoire, seeking, as Móller, Pedersen, Cuny, and Kuryfowicz had
done before them, to adduce further proof to the basic correctness of Saussure's theories
(cf. Kodzu's (1939) review article on this subject). As Polomé in his formidable survey of
the history of the laryngeal theory has shown (Polomé 1965: 14ff.), the work of Kuryto-
wicz, Benveniste, Couvreur and others was first met with a lack of sympathy and com-
prehension similar to the fate of the Mémoire during Saussure's life-time, namely, that the
hypotheses were characterized as ‘constructed’, ‘aprioristic’, ‘too audacious to be true',
and the like. History appears to be repeating itself.

213
2.0. Concluding remarks

In this paper, it has not been my intention to trace the history of the laryngeal theory; this
has been done already very competently by others (e.g., Polomé 1965; Szemerényi 1973;
Jonsson 1978). Nor did I mean to indicate the role that Saussure's proposals concerning
the structure of the Indo-European vowel system have played in the development of this
theory, by simply suggesting that he had given the initial impetus to research undertaken
by others. The reason for this is simple: I believe that Saussure had much more to offer to
historical linguistics than an initial suggestion which was subsequently proved to be right,
although this may be no small contribution by itself. Indeed, we all have witnessed that
with the advent of the so-called Chomskyan ‘paradigm’ hundreds of linguists and would-be-
linguists have been engaged in tossing out ‘theories’ almost daily, frequently withdrawing
them a few weeks later themselves before others had time to refute them.

In Saussure’s case we have to do with something much more profound and valuable. Apart
from the fact never to be overlooked that his argument was based on a considerable amount
of verifiable data, Saussure presented his theory of the Proto-Indo-European vowel system
in a highly organized fashion, proceding with a step-by-step interpretation of the material
together with his reasoning about the data. Most importantly, he put forward a theory
arrived at not merely in a wholly deductive fashion deriving from the premise that the
vowel system of any language must be a complete system of relationships, but at the same
time by adducing inductively, as much as was possible at the time, anything that could
lead to substantiating the veracity of his general assumption. That he produced a hypothesis
of consequence was therefore not simply the strike of genius — which is true nevertheless
of course, but the result of a considerable familiarity with the available data, the toois of
the craft, and a particularly clear mind. This last-mentioned criterion may be the key to
Saussure’s success, distinguishing himself from the great majority of his contemporaries
and, we may add, followers. Saussure was very conscious about this, and critical especial-
ly of the linguists of his day who evidently lacked the capacity which he regarded as vital
to the advancement of the field. In an unpublished draft of a letter probably dating from
the 1890s, when he was planning on writing a book on the principles of linguistic science,
Saussure stated:

"Je veux malgré cela résumer quel est pour moi l'exact état des preuves, car ce que je
déteste chez tous les Germains comme [aussi Holger] Pedersen, c'est la maniére subreptice
d'amener la preuve, et de ne jamais la formuler, comme si la preuve de leurs réflexions
les dispensait de mettre à nu leur opération logique". (Quoted from Watkins [1978: 65])

What we may learn from Saussure's example for historical linguistics as well as for the
study of language in general is the following:

(a) The analyst must lay bare, to himself and others, the entire procedere of his argument;
his approach must be clearly presented and well reasoned.

(b) Historical-comparative linguistics is not a field of mere data gathering and ordering, a
barren positivistic enterprise as has been common at least since the 1870s, but a subject
which can only advance significantly if it admits the setting-up of hypotheses which must
be substantiated by subsequent research.

(c) Language in general is an ordered whole; as a result, the tools of analysis must be used
on the assumption of the basic systematicity of the data.

214

These and other ideas have become familiar to us through the Cours de linguistique géné-
rale, although the manner in which it was compiled has left the impression on three genera-
tions of linguists that Saussure was holding historical linguistics in low esteem.? However, a
close reading of Saussure’s Mémoire sur le système primitif des voyelles makes these points
very clear, indeed offering us in hindsight a proof for the desirability of a sound combina-
tion of theory and empirical evidence, of deduction and induction, in historical linguistic
research. That his Mémoire has still today not yet lost its original lustre may be taken from
recent evaluations of this work (e.g., Vincenzi 1978; Watkins 1978), and a recent announ-
cement (Gmiir 1980: 4) of a comprehensive dissertation devoted to this pioneering work.

Notes

l Saussure must have drawn some belated satisfaction from Brugmann’s observation (1966 [1909]:

560) according to which Osthoff’s monograph (1881) constituted in effect a step back (‘‘Riick-
schritt”) vis-à-vis the Mémoire.

Curtius was not referring to the Mémoire, whose rigor might well have exceeded his tastes and com-
prehension, but to an article of 1884, “Une loi rythmique de la langue grecque”, of which Curtius
had taken notice, one year before his death, which had probably been precipitated by the ‘war of
monographs’ between him and his former students concerning the sound laws and the analogy
principle during 1885. — The quotation is taken from Redard (1978: 36).

Watkins (1978: 67), who quotes the second half of the sentence, argues that the phrase ‘structure
considérée en elle-méme' italiziced in the citation constitutes “la véritable source” of a part of the
famous statement at the end of the Cours, for which Engler was at a loss to find a textual basis in the
critical edition (p. 515): la linguistique a pour unique et véritable objet la langue envisagée en elle-
même et pour elle-même (Cours, p. 317). This is undoubtedly an overinterpretation, but it merits
further investigation in the light of the entire context of Saussure’s argument, both in the Mémoire
and elsewhere in his writings, published as well as unpublished.

Source: Lehmann (1967: 233). I have ventured to introduce a couple of Verschlimmbesserungen
into the translation on the basis of my understanding of the French original (Saussure 1879: 51—
52), where, among other things, certain phrases are in italics, but which Lehmann presented in
regular type. (The emphasis in the original appears to me to underscore the forcefulness of Saussure’s
manner of argumentation.)

When Saussure speaks of “i et e passent de l'état symphthongue à l’&tat autophthonge” (1879: 8) --
note also the use of “état” in the present context! — it seems to me that he is using a terminology
(meaning "functioning as a consonant" and "functioning as a vowel", correspondingly) created by
the French phonetician A. Dufriche Desgenettes (cf. Koerner 1976: 227—228).

Interestingly enough, the great English phonetician Henry Sweet (1845 —1912), in his review of the
Mémoire, calls the sonant coefficients A and Q *consonants' or ‘consonantal elements’ (Sweet 1880:
160). One may wonder whether it would be correct to say with Jonsson (1978: 2, n.1) that “we
have to do [here] with an anachronism or a misinterpretation under the influence of the later con-
sonantal theory", without indicating what he meant by ‘later’. However, Sweet (1880: 158) refers to
Möller’s (1879) review of Kluge (1879), where in fact the consonantal hypothesis was first enunciat-
ed (see 1.2. below for details).

Mayrhofer (1981: 28-29, note 81) adduces evidence for his suggestion that the term ‘schwa’ was
first introduced by August Fick; since he refers only indirectly to the possibly first appearance of the
term in print, reference is made here to Fick’s paper (1879).

Pedersen had been under the influence of Möller’s views from very early on in his scholarly activity
(from 1893 onwards, if not earlier); cf. the bibliography of his major writings in H. Pedersen (1983:
xxiii-XXX). For details on Pedersen's contribution to the laryngeal theory, cf. Szemerenyi (1973:
11-12 and 15).

Thus, ever since the availability of Rudolf Engler's critical edition of the Cours (1967-1974), it
should have become clear to everybody that the various statements in the text as compiled by Charles
Bally and Albert Sechehaye concerning the clash between synchrony and diachrony for instance
have little foundation in Saussure's lectures as recorded by his students.

215

References

Arlotto, Anthony, 1972. Introduction to Historical Linguistics (Boston: Houghton Mifflin Co.).

Benveniste, Emile, 1935. Origines de la formation des noms en indo-européen, Vol. 1 (Paris: A. Maison-
neuve). [Actually published in 1936.]

Brugmann, Karl, 1876. “Nasalis sonans in der indogermanischen Grundsprache”, (Curtius’) Studien der
griechischen und lateinischen Grammatik 9: 285—338. [Abridged English translation in Lehmann
(1967: 191-196).}

Brugmann, Karl, 1879a. “Review of Saussure 1879”, Literarisches Zentralblatt für Deutschland 29: coll.
713—114.

Brugmann, Karl, 1879b. “Nachschrift [of June 1879]". Morphologische Untersuchungen auf dem Ge-
biete der indogermanischen Sprachen 2: 161—162.

Brugmann, Kari, 1909. “Hermann Osthoff”, Indogermanische Forschungen 24: 218-223 [Quoted
after the reprint in Portraits of Linguists, edited by Thomas A. Sebeok, vol. 1, 1966 (Bloomington
& London: Indiana University Press), 555—562.].

Brogyanyi, Bela, 1983. “A Few Remarks on the Origin of the Phrase ‘où tout se tient
Linguistica 10: 143—147.

Couvreur, Walter, 1937. De hettitische h: Een bijdrage tot de studie van het indo-europeesche vocalisme
(Louvain: Muséon). [French summary, 381—391.]

Couvreur, Walter, 1939. “Le hittite et la doctrine de F. de Saussure”, Revue hittite et asianique 5: 132-
141.

Cuny, Albert, 1912. “Notes de phonétique historique: Indo-européen et sémitique", Revue de Phonéti-
que 2: 101—132.

Fick, August, 1879. “Schwa indogermanicum”, Beitrige zur Kunde der indogermanischen Sprachen
3: 157-165. |

Fick, August, 1880. “Review of Saussure 1879”, Göttingische gelehrte Anzeigen, Stück 14 (7. April
1880): 417-439.

Gmür, Remo, 1980. Das Mémoire von F. de Saussure (Bern: Institut für Sprachwissenschaft, Universität
Bern).

Havet, Louis, 1879. “Review article on Saussure 1879”, Journal de Genève No. 47: [50e année] (Sup-
plement) 1-2 (25 Feb. 1879). [Cited after reprint in Cahiers Ferdinand de Saussure 32: 103—122
(1978).] |

Hirt, Herman, 1900. Der indogermanische Ablaut, vornehmlich in seinem Verhältnis zur Betonung
(Straßburg: Karl J. Trübner).

Hirt, Herman, 1927. Indogermanische Grammatik, Vol. 1 (Heidelberg: C. Winter).

Hjelmslev, Louis, 1970. Language: An introduction [Transl. from the Danish by Francis J. Whitfield]
(Madison-Milwaukee-London: Univ. of Wisconsin Press).

Hrozný, Bedřich, 1915. “Die Lösung des hethitischen Problems: Ein vorläufiger Bericht”, Mitteilungen
der Deutschen Orient-Gesellschaft zu Berlin 56: 17-50.

Hrozný, Bedfich, 1916-1917. Die Sprache der Hethiter: Ihr Bau und ihre Zugehörigkeit zum indoger-
manischen Sprachstamm: Ein Entzifferungsversuch (Leipzig: J. C. Hinrichs).

Hübschmann, Heinrich, 1885. Das indogermanische Vocalsystem (Straßburg: Karl J. Trübner).

Jonsson, Hans, 1978. The Laryngeal Theory: A critical survey (Lund: C.W.K. Gleerup).

Kluge, Friedrich, 1879. Beiträge zur Geschichte der germanischen Conjugation (Straßburg: Karl J.
Trübner). |

Kodzu, Harushige, 1939. “Lu-o-go boin henka no kenkyu to ‘laryngales’ no hakken” [Vocalic change of
Indo-European and the discovery of laryngeal sounds], Gengo Kenkyu 3: 53—76.

Koerner, E.F.K., 1976. “A Minor Figure in 19th-Century Linguistics: A. Dufriche Desgenettes”, Pho-
netica 33.222—231. [Repr. in Koerner, Toward a Historiography of Linguistics (Amsterdam: John
Benjamins, 1978), 127—136.]

Kruszewski, Mikotaj, 1880. *'Novejtija otkrytija v oblasti ario-evropejskogo vokalizma [Recent discov-
eries in the field of Indo-European vocalism]”, Russkij Filologiteskij Vestnik (Warsaw) 4: 33—45.
[A review article on Brugmann (1876) and, in particular, Saussure (1879); Italian translation in
Saussure 1978: 441—450.]

Kuryłowicz, Jerzy, 1927. “2 indo-européen et h hittite”, Symbolae Grammaticae in honorem Ioannis
Rozwadowski, vol. 1 (Krakow: Drukarnia Uniwersytetu Jagiellońskiego), 94—104.

aan

, Historiographia

216

Lehmann, Winfred Philipp (ed.), 1967. A Reader in Nineteenth-Century Historical Indo-European Lin-
guistics (Bloomington & London: Indiana University Press).

Mayrhofer, Manfred, 1981. Nach hundert Jahren: Ferdinand de Saussures Frühwerk und seine Rezep-
tion durch die heutige Indogermanistik (Heidelberg: C. Winter).

Mayrhofer, Manfred, 1983. “Sanskrit und die Sprachen Alteuropas: Zwei Jahrhunderte des Widerspiels
von Entdeckungen und Irrtümern", Nachrichten der Akademie der Wissenschaften in Göttingen 1:
Philologisch-historische Klasse 5: 121--154 (Göttingen: Vandenhoeck & Ruprecht).

Meillet, Antoine, 1903. Introduction à l'étude comparative des langues indo-européennes (Paris: Hachet-
te).

Meyer, Gustav, 1880. Griechische Grammatik (Leipzig: Breitkopf & Hartel). [3rd reversed & enlarged
edition, 1896.]

Möller, Hermann, 1879. “Review of Kluge 1879”, Englische Studien 3: 148—164.

Möller, Hermann, 1880. “Zur Conjugation: kunpa und das t-praeteritum. Excurs: Die Entstehung deso”,
Beiträge zur Geschichte der deutschen Sprache und Literatur 7: 482—534. [“Excurs” 492—534.].

Möller, Hermann, 1906. Semitisch und Indogermanisch 1: Konsonanten (Copenhagen: H. Hagerup).

Möller, Hermann, 1908. “Die gemein-indogermanisch-semitischen Worttypen der zwei- und dreikonso-
nantigen Wurzeln und die indogermanisch-semitischen vokalischen Entsprechungen”, Zeitschrift für
vergleichende Sprachforschung 42: 174-191.

Möller, Hermann, 1911. Vergleichendes indogermanisch-semitisches Wörterbuch (Göttingen: Vanden-
hoeck & Ruprecht). [See esp. Preface xviii—xxi.].

Möller, Hermann, 1917. Die semitisch-vorindogermanischen laryngalen Konsonanten (= Mémoires de
l'Académie Royale des Sciences et des Lettres de Danemark, Te série, 4:1.) (Copenhagen: F.A. Høst
& Son).

Osthoff, Hermann, 1879. “Kleine beiträge zur deklinationslehre der indogermanischen sprachen 2”,
Morphologische Untersuchungen auf dem Gebiete der indogermanischen Sprachen 2: 1—147.

Osthoff, Hermann, 1881. “Die Tiefstufe im indogermanischen Vocalismus”. Morphologische Untersu-
chungen auf dem Gebiete der indogermanischen Sprachen 4:1.-406.

Paul, Hermann, 1880. Principien der Sprachgeschichte (Halle/S.: M. Niemeyer). [5th reversed edition
1920.]

Pedersen, Holger, 1907. “Die indogermanisch-semitische Hypothese und die indogermanische Lautleh-
re”, Indogermanische Forschungen 22: 341—365.

Pedersen, Holger, 1983. A Glance at the History of Linguistics, with particular regard to the historical
study of phonology, edited with an introduction by Konrad Koerner (Amsterdam & Philadelphia:
John Benjamins).

Petersen, Walter, 1939. “Hittite k and Saussure’s Doctrine of the Long Vowels”, Journal of the Ameri-
can Oriental Society 59: 175--199.

Polomé, Edgar, 1965. “The Laryngeal Theory so far: A critical bibliographical survey”, in: Evidence
for Laryngeals, edited by Werner Winter (London/The Hague: Mouton), 8—78. [Bibliography 44—
78.]

Redard, Georges, 1976. “Ferdinand de Saussure et Louis Havet”, Bulletin de la Société de Linguistique
de Paris 71: 1.313—349. [Publication of correspondence between Saussure and Havet.]

Redard, Georges, 1978. '*Deux Saussure?", Cahiers Ferdinand de Saussure 32: 27-241.

Regnaud, Paul, 1889. Esquisses du véritable système primitif des voyelles dans les langues d’origine indo-
européenne (Paris: E. Leroux).

Regnaud, Paul, 1890. “Le véritable système vocalique indo-européen: Preuves et déductions nouvelles”,
Revue de Linguistique et de Philologie comparée 23: 63—89, 183—187.

Regnaud, Paul, 1891. Observations critiques sur le système de Monsieur de Saussure (Gray/Haute Saône:
Bouffaut Frères).

Rhys, (Sir) John, 1879. "Review of Saussure 1879", The Academy (London, 27 Sept. 1879), 234—235.

Saussure, Ferdinand de, 1879. Mémoire sur le systéme primitif des voyelles dans les langues indo-euro-
péennes (Leipzig: B. G. Teubner). [Reprinted Hildesheim: G. Olms, 1968; actually published in
December 1878.]

Saussure, Ferdinand de, 1891. "Contribution à l'histoire des aspirées sourdes", Bulletin de la Société de
Linguistique de Paris 7: 143 (1892). [Summary of a paper presented at the ‘Séance du 6 juin 1891’.
Reprinted in Saussure 1922: 603.1.

217

Saussure, Ferdinand de, 1916. Cours de linguistique générale, edited by Charles Bally & Albert Seche-
haye, with the collaboration of Albert Riedlinger (Lausanne & Paris: Payot).

Saussure, Ferdinand de, 1931. Cours de linguistique générale, third corrected edition (Paris: Payot).
(All other editions follow this one.]

Saussure, Ferdinand de, 1922. Recueil des publications scientifiques, edited by Charles Bally & Léopold
Gautier (Geneva: Editions Sonor; Lausanne: Payot; Heidelberg: C. Winter). [Reprinted Geneva:
Slatkine, 1970.]

Saussure, Ferdinand de, 1967 —1968, 1974. Cours de linguistique générale, édition critique par Rudolf
Engler (Wiesbaden: O. Harrassowitz).

Saussure, Ferdinand de, 1977. “Memuar o pervonatal’noj sisteme glasnych v indoevropejskich jazykach”,
Trudy po jazykoznaniju by F. de Saussure, edited by A.A. Cholodovit (Moscow: Progress), 302—
562. [= Russian transl. of Saussure 1879.]

Saussure, Ferdinand de, 1978. Saggio sul vocalismo indoeuropeo. Edizione italiana a cura di Giuseppe
Carlo Vincenzi (Bologna: Cooperativa Libraria Universitaria Editrice). [= Italian transl. of Saussure
1879.]

Schleicher, August, 1861-1862. Compendium der vergleichenden Grammatik der indogermanischen
Sprachen: Kurzer Abriß einer Laut- und Formenlehre der indogermanischen Ursprache (Weimar:
H. Böhlau). [4th edition prepared by August Leskien & Johannes Schmidt, 1876.]

Schmitt-Brand, Robert, 1967. Die Entwicklung des indogermanischen Vokalsystems: Versuch einer in-
neren Rekonstruktion (Heidelberg: Julius Groos).

Streitberg, Wilhelm, 1914. "Ferdinand de Saussure"', Indogermanisches Jahrbuch 2: 203—213. (Reprint-
ed in Portraits of Linguists, edited by Thomas A. Sebeok, vol. 2, 1966: 100—110 (Bloomington &
London: Indiana Univ. Press.)

Sweet, Henry, 1880. "Recent Investigations of the Indogermanic Vowel-System", Transactions of the
Philological Society 1880—1881: 155—162. (Oxford: Clarendon). {Reprinted in H. Sweet, Collected
Papers, edited by Henry C. Wyld 1913: 141—147.]

Szemerényi, Oswald, 1973. “La théorie des laryngales de Saussure à Kuryłowicz et à Benveniste: Essai
de réévaluation”, Bulletin de la Société de Linguistique de Paris 68: 1-25.[Translated from the
English by Michel Lejeune.]

Szemerényi, Oswald, 1980. Einführung in die vergleichende Sprachwissenschaft, second reversed edition
(Darmstadt: Wissenschaftliche Buchgesellschaft). [First edition 1971.]

Vallini, Cristina, 1969. Problemi de metodo in Ferdinand de Saussure indoeuropeista (Pisa: P. Mariotti).

Vincenzi, Giuseppe Carlo, 1976. “‘‘Sistema’ e ‘fonema’ nel primo Saussure”, Studi italiani di Linguistica
teorica ed applicata 5: 229-251.

Vincenzi, Giuseppe Carlo, 1978. "Introduzione del Curatore", in: Saggio sul vocalisme indoeuropeo
(Bologna: Cooperativa Libraria Universitaria Editrice), xix—xciv.

Watkins, Calvert, 1978. "Remarques sur la méthode de Ferdinand de Saussure comparatiste", Cahiers
Ferdinand de Saussure 32: 59-68. [Appendice, 68—69.]

Archaic ablaut patterns in the Vedic verb

Frederik Kortlandt (University of Leiden)

1. The first singular active form of the Vedic sigmatic aorist injunctive does not take
vrddhi. This is a remarkable archaism which has not been sufficiently appreciated.

2. In his article on the “proterodynamic” root present Insler calls attention to the fact that
“the system of proterodynamic present inflection reflected in Vedic forms is nearly identi-
cal to the oldest system of Vedic sigmatic aorist inflection” (1972: 56). “It is only when
we compare the act. indic.-inj. of proterodynamic root presents that the complete paral-
lelism breaks down" (Insler 1972: 57). The active forms of the sigmatic aorist have length-
ened grade vocalism throughout the whole paradigm and do not show the expected alterna-
tion between lengthened grade in the singular and full grade in the plural which is found in
tasti, táksati. We must therefore ask the question: ‘which paradigm seems to continue the
original ablaut relationship?" (Insler 1972: 58).

3. The obvious explanation is that the active paradigm of the sigmatic aorist ‘has partici-
pated in the same sort of leveling of vocalism observed in act. root aorists of the type
dkar, dkarma, dkarta" (Insler 1972: 58). Insler rejects this view because the lengthened
grade vocalism was extended to the third plural form of the sigmatic aorist, whereas the
corresponding form of the root aorist maintains the original zero grade, e.g. dkran. The
argument does not hold because the ending of the root aorist was -an < *-ent, whereas the
sigmatic form ended in *-sat < *-snt. The ending *-at was replaced with -ur, as it was in the
injunctive taksur and in the reduplicated imperfect. The retention of the ablaut contrast in
the paradigm of tasti and the extension of the lengthened grade to the third plural form of
the sigmatic aorist fit “the general tendencies of the Vedic verb system to characterize act.
athematic present inflection by ablaut differences, but to mark act. athematic aorist in-
flection by the predominant absence of any alternating vocalism" (Insler 1972: 61).

4. Lengthened grade vocalism was generalized in the active paradigm of the sigmatic aorist
indicative, but not in the injunctive, which betrays the original distribution of the ablaut
grades. It is noteworthy that the original distribution was already indicated by Wackernagel
in his Old Indic grammar (1896: 68): the lengthened grade spread from the monosyllabic
second and third singular forms to the rest of the paradigm. The archaic character of this
distribution is supported by the Balto-Slavic evidence (cf. Kortlandt 1975, App. E, and
1984, sections 1.3 and 1.4). It is also clear from the Vedic material.

5. The first singular indicative has lengthened grade in RV. ajaisam, apraksam, abharsam,
ayamsam, asparsam, aharsam, akanisam, akarisam, acarisam, asanisam, and ambiguous

220

vocalism in aydsam. It has full grade in akramisam and in the analogic forms akramim and
asamsisam. The first person singular injunctive ‘has full grade in VS.TS.TB. jesam, TS.KS.
TB.JB. yosam, and RV. stosam, vadhim, and lengthened grade in the analogic form ravisam
(ru- ‘break’).

6. Following Hoffmann, Narten interprets jesam and first person plural RV. jesma as
precative forms (1964: 120). The reason for this interpretation is evidently the absence
of lengthened grade (cf. Hoffmann 1967a: 254). The functional evidence for the interpre-
tation as precative (Hoffmann 1967b: 32f) or subjunctive (Insler 1975: 15%) is very weak,
while the formal objections against it are prohibitive. It is therefore preferable to retain the
traditional view that these forms are what they look like: full grade injunctive forms, which
were interchangeable with the corresponding subjunctive in certain contexts and which
could be interpreted as precative when the latter category became common.

7. Narten assumes that the injunctive forms yosam and stosam took their vocalism from
the subjunctive (1964: 213, 277). The model for this analogic development is lacking,
however, because the subjunctive ending was -ani, not -am. Hoffmann attributes the alleged
substitution of the injunctive ending -am for the earlier subjunctive ending -4 to the in-
fluence of the second singular imperative: “Das Bestreben, den Konjunktivausgang -4 von
dem durch Auslautsdehnung gleichlautend gewordenen Imperativausgang zu sondern, hat
das Ausweichen zu -am, wodurch die 1. Person deutlich gekennzeichnet wurde, gefördert”
(1967a: 248). I find such influence highly improbable. The use of the first person singular
injunctive for the subjunctive must be explained from the meaning of the forms. Note that
standard British English offers an exact parallel in the use of ‘I shall’ where other persons
“will’.!

8. The indicative has lengthened grade in RV. third dual asvarstam, first plural ajaisma,
abhaisma, atarisma, second plural achanta, third plural achantsur, abhaisur, atarisur,
apavisur, amädisur, aranisur, aravisur, avadisur, asavisur, and ambiguous vocalism in ayäsur,
arajisur, and avisur. It has full grade in third dual amanthistam, first plural agrabhisma,
third plural ataksisur, adhanvisur, anartisur, amandisur, all of which have a root in a double
consonant (cf. grbhita- < *grbhH-itd-). It has zero grade in amatsur, anindisur, and aksisur
(nas- ‘attain’).

9. The injunctive has full grade in Rgveda second dual avistam, kramistam, gamistam,
canistam, cayistam, mardhistam, yodhistam, vadhistam, snathistam, third dual avistam,
first plural jesma, sramisma, second plural avista(na), grabhista, ranistana, vadhista(na),
snathistana, zero grade in himsista, and ambiguous vocalism in third plural dhasur, hasur.
It has lengthened grade in second dual yaustam (ApMB. yostam), taristam, second plural
naista (ApSS. yosta), third plural yausur, jarisur, and in the analogic form second dual
yavistam (yu- *unite"). Note that the difference between first plural sramisma and atarisma
parallels the one between jesma and ajaisma.

10. One may wonder if the ablaut difference between the indicative and the injunctive is
also found in the asigmatic aorist. It has long been noticed that the third plural middle in-
dicative forms dkrata and drata correspond with the injunctive forms kranta and ranta
(Meillet 1920: 203, 205). The archaic character of this distribution is supported by the

221

isolated third plural injunctive forms nasan and nasanta, which correspond with indicative
aksisur (for asur replacing "asat) and asata. Hoffmann’s conjecture that the initial n- of the
injunctive is of secondary origin (1957: 124f) does not explain why it is limited to the
third person plural forms, cf. third singular middle asta. As in the case of the sigmatic
aorist, it is probable that the vocalic alternation was eliminated in the indicative paradigm.
This must have occurred at a much earlier stage, however, because it affected the form
which was to yield dsthur. The full grade injunctive ending -anta survived in the paradigm
of the subjunctive, which shared the thematic vowel. There is a trace of the original distri-
bution in Homer ravvraı, ravvovro.

11. As I indicated above (section 3), the third plural ending -ur replaced earlier *at < *unt,
not -an < *-ent. Since the optative ends in -yur, the original form must have had zero grade
both in the suffix and in the ending. This suggests that it had full grade in the root.

12. Hoffmann has argued that the root aorist optative had fixed stress on the root (1968).
His proposal offers a straightforward explanation for third plural Latin velint, Gothic
wileina, and OCS. velet», but not for the remarkable alternation which the latter language
shows between second plural xo¥tete, dovsljete and third plural xotets, dovbletr. It ap-
pears that the third person plural form differed from the other persons in the original
paradigm. This enables us to remove the unlikely assumption that the root aorist differed
from the root present in the accentuation of the optative.

13. Insler connects the type dheyam with the type gaméyam, the two being in comple-
mentary distribution (1975: 15). His explanation falters on two points. First, it requires
the previous existence of both *dheyam and *dhayam, of which the attested form repre-
sents a blending. It is highly improbable that neither of the earlier forms would have sur-
vived because both were supported by other paradigms, while the alleged blending created
a new type. Second, the motivation for the spread of the new vocalism to the third person
forms is very weak. The long chain of analogic changes which Insler’s theory requires is
too complicated to be credible.

14. Thus, I arrive at the following reconstruction of the Proto-Indo-European active root
optative:

Ist sg. dhHjieH;m pl. dhHjiH jme

2nd dhHjieH}s dhH jiH jte

3rd — dhHjieH jt dheH jiH jnt
After Sievers' law and the loss of tautosyllabic laryngeals this paradigm turned into the
following:

Ist sg. dhiyam pl. dhima
2nd dhiyas dhita
3rd dhiyat dha?i?at

The generalization of *dha?- and the substitution of -ur for *-at yielded first singular
dheyám, third plural dheyur.?

222

15. The isolated first person plural middle optative form nasimahi (three times) next to
asimahi (five times) suggests that this paradigm also contained a form with full grade in the
root. Since the initial n- is lacking elsewhere in the middle optative and indicative para-
digms, it was probably taken from the unattested third person plural optative form.

16. The accentual mobility in the paradigm of the optative is reminiscent of the one in the
reduplicated present, where third plural bíbhrati and dadhati have both initial stress and
zero grade in the root and in the ending. Thus, I reconstruct PIE *dhedhH nti ‘they put’.
It follows that the third person plural form does not have the same origin as the other
forms of the paradigm.

17. The reduplicating syllable da- of didhami replaces earlier di-, which is preserved in
riümu and in the desiderative present didhisami. It is difficult to agree with Leumann's
view that da- was taken from the perfect (1952: 27) because the motivation for such an
analogic development was very weak. More probably, the paradigm of the present contained
a form with da- from the very outset. This must have been the third person plural form. In
my view, PIE *dhi- was simply the pretonic (zero grade) variant of *dhe- before a double
consonant, cf. mirvnui, miovpes, Czech čtvrtý < *č»tvrtyj ‘fourth’, OCS Xs/s « *Ssdi»
‘went’.

18. The third person plural forms yánti, kranta, dheyur, and dadhati have in common that
the initial syllable contains a full grade vowel. They have the same vocalism as the partici-
ples yánt-, kránt-, dádhat-. It is therefore probable that the form in -nti represents the
original nominative plural form of the participle. The plural ending -i is also found in the
Proto-Indo-European pronominal inflection: nom. "fo-i, gen. *fo-i-s-om, dat. *to-i-mus,
abl. *to-i-os, inst. *to-i-bhi, loc. *to-i-su. It follows from this point of view that the second-
ary ending *-nf was created on the analogy of the singular forms, where the primary -i had
a different origin ?

Notes

During my stay in Dublin, Dr. Patrick Sims-Williams told me that when an Irish friend asked him in
front of an open door: “Will I go first?", the only reasonable answer to him would be: “I don't
know”. Compare in this connection RV. VII 86.2 kadá nv äntär värıne bhuvani ... kadä mrlikäm
sumdna abhi khyam ‘When will I be inside Varuna? When shall I, cheerful, perceive his mercy?’ Also
X 27.1 ásat sú me jaritah sábhivegó, yát sunvaté ydjamänäya $iksam "That will be my excitement,
singer, that I shall be helpful to the pressing sacrifier'. In X 28.5 kathá ta etád ahám a ciketam ‘How
shall I understand this (word) of yours?' the substitution of the subjunctive for the injunctive would
yield a quite different shade of meaning: it would shift the responsibility from the singer to Indra.
Forms of the type dheyäm are always trisyllabic in the Rgveda. This fits the explanation advanced
here.

In my view, the plural ending -i is of Indo-Uralic origin. It can be identified with the Fennic and
Northern Samoyed oblique plural suffix -i-, e.g. Finnish talo ‘house’, pl. (alot, taloi-. It is also found
as a plural object marker in the Northern Samoyed objective conjugation, e.g. Yurak mada-i-n ‘I
(did) cut (more than two things)’, cf. Finnish pala-n ‘I burn (intr.)’, Lappish puolam < *palak-mi.

223

References

Hoffmann, Karl, 1957. “Zur vedischen Verbalflexion”, Münchener Studien zur Sprachwissenschaft 22:
121-137.

Hoffmann, Karl, 1967a. Der Injunktiv im Veda (Heidelberg: Harrassowitz).

Hoffmann, Karl, 1967b. “Der vedische Prekativtyp yesam, jesma", Münchener Studien zur Sprachwis-
senschaft 20: 25-37.

Hoffmann, Karl, 1968. “Zum Optativ des indogermanischen Wurzelaorists”, Pratidänam 3-8 (The
Hague: Mouton).

Insler, Stanley, 1972. “On proterodynamic root present inflection”, Münchener Studien zur Sprachwis-
senschaft 30: 55-64.

Insler, Stanley, 1975. “The Vedic type dheyam”, Die Sprache 21: 1—22.

Kortlandt, Frederik, 1975. Slavic accentuation (Lisse: Peter de Ridder Press).

Kortlandt, Frederik, 1984. “Long vowels in Balto-Slavic”, Baltistica 21: 112—124.

Leumann, Manu, 1952. Morphologische Neuerungen im altindischen Verbalsystem (Amsterdam: KNAW).

Meillet, Antoine, 1920. “Sur le rythme quantitatif de la langue védique”, Mémoires de la Société de Lin-
guistique de Paris 21: 193-207.

Narten, Johanna, 1964. Die sigmatischen Aoriste im Veda (Wiesbaden: Harrassowitz).

Wackernagel, Jacob, 1896. Altindische Grammatik 1: Lautlehre (Göttingen: Vandenhoeck & Ruprecht).

Theoretical views affecting successive reconstructions of
the phonological system of Proto-Indo-European

Winfred P. Lehmann (University of Texas)

0. Introduction

The phonological system proposed for a language, even for a reconstructed language, may
seem to rest on the securest evidence available for that language, on evidence far more
secure than that available for positing its syntactic system or its morphology. The changing
views on the phonological system of Proto-Indo-European raise doubts concerning such an
assumption. Within the period of scarcely more than one century, various phonological
systems have been proposed for Proto-Indo-European. All have been based on essentially
the same data. Only the Anatolian languages provided significant additional data after their
discovery in the first decade of this century, data especially crucial to those who place little
credence in internal reconstruction. While the data are largely unchanged, the successive
systems proposed illuminate the changing theoretical views of linguists during the past
hundred years. Since historical theory in its development has been of special interest to the
scholar honored by this volume, I take this opportunity to review selected innovations in
that theory with reference to their impact on the proposed phonological system of Proto-
Indo-European.

Examination of the posited systems in turn provides perspectives on those in handbooks of
the past, perspectives which may also be useful for evaluating phonological constructs pro-
posed today and in the future. In presenting the systems proposed by selected Indo-Euro-
peanists, I am primarily concerned with the theoretical views on which they were based,
rather than with an optimum reconstruction of the phonology of Proto-Indo-European.

1. Schleicher and his system

The first system presented here is that of August Schleicher (1871: 10). Earlier linguists
did not propose a separate Indo-European system, assuming instead that that of Sanskrit
applied too for the parent language. A glance at Schleicher’s system indicates that even it is
heavily influenced by the Sanskrit system. Schleicher does not posit e or o. He does not
include /. There are no labiovelars.

226

Table 1
Schleicher’s phonological system
Consonanten Vocale
momentane laute dauerlaute
nicht aspirierte aspiratae spiranten nasale r-laute
stumm tôn. tönend. stumm tônend. tön. tön.
gutt. k g gh a) aa da
pal. j i\}ai di
lingu. r au d
dent. t d dh S n
lab. p b bh y m u

In keeping with our focus on the theories underlying posited systems, I limit myself to a
few essentials rather than recalling other highly interesting remarks of Schleicher concern-
ing his own views in comparison with those of his predecessors and colleagues.

First, Schleicher assumed an earlier system, which we might label pre-Indo-European. Such
an assumption is admirable; it has been far too often neglected by subsequent Indo-Euro-
peanists. For Schleicher, the system of Table 1 developed from an earlier one lacking
voiced aspirates and diphthongs with @. An even earlier system, the most original (“ur-
spriinglichsten”) lacked all diphthongs. His system of 1871 is that of the language ‘shortly
before the first division.’

Second, Schleicher clearly posited a symmetrical set of sounds. From it we may conclude
that he aimed to propose a system, not simply a list of individual sounds.

His aim to posit a system apparently had further consequences, of which I cite two. One
concerns the absence of laryngeals. In a note he states that the ‘aleph or hamza of the
Semites’ was found before initial vowels. The omission of this element from his roster gives
us insights into his theoretical views. First, he did not regard morphological structure in
proposing his system. Further, the aleph was disregarded as an element not in direct
contrast with other consonants and accordingly extra-systematic. — As an aside we may
wonder whether his suggestion of a laryngeal in the system may not have had important
consequences. Saussure, whose monograph is dated eight years later, surely knew Schlei-
cher's Compendium. So must have Móller, who identified Saussure’s (1879) coefficients
as laryngeals in the year 1880.

A second consequence of Schleicher’s systematization is his positing of b. Schleicher states
that he knows no completely certain examples of this sound, adding that Grassmann made
it highly probable that b was not present in the proto-language, at least not initially (1871:
160). Systematization apparently led Schleicher to include it nonetheless.

2. The neogrammarians

The following decade produced an explosion of contributions on the phonological system.
The period abounded with new ideas and argumentation. Already in Schleicher’s day two

227

rival camps disputed the importance of adhering strictly to sound laws. Among those in
favor of doing so Schleicher lists G. Curtius in Leipzig, Corssen in Berlin and himself. (In
spite of this classification, Curtius comes down to us as superseded father of the neo-
grammarians.) Among modifications of Schleicher's system introduced by adherents to
sound laws are the vowels e and o; unshakeable evidence was provided from Sanskrit in the
so-called law of palatals.

Insistence on sound laws led to concern with all residues, also termed exceptions to sound
laws. Verner's explanation of the third set of exceptions to Grimm's rules, exceptions well-
known for fifty years, led to the assumption of suprasegmentals, and to clarification of the
Indo-European system of ablaut. In short order, Schleicher's system was completely
superseded. The course of the modification and their bases are well-presented in Fritz
Bechtel (1892). In this survey Bechtel includes publications to January 1890, less than
20 years from the publication date of Schleicher’s third edition of the Compendium.

Examination of Brugmann's phonological system discloses changes that had been adopted
in these few years. I provide the system as presented in Brugmann (1904).

Table 2
Brugmann's phonological system
1) Vokale. ii, ut, e é, 0 6, aa, 0 (‘Schwa?). — i, u. — Fallende Diphthonge: ei oi ai
oi, ĉi Oi ài; eu ou au ou, éu du du.
Nasale. m (labial), n (dental), fì (palatal), y (velar). Silbisch: m m, n n, ñ fi, nn.
Liquidae. r,1. Sonantisch: r z, 11.

2) Verschlußlaute (Explosivae).

p ph b bh (labial, genauer bilabial),

t th d dh (dental, genauer vermutlich alveolar),
k kh g gh (palatal),

q qh g_ gh (reinvelar),

q- q*hg* gh (labiovelar).

3) Reibelaute (Spiranten). s und sh, z und zh; p und ph, è und ðh. Dazu vielleicht j.

Clearly the elements in the system have been vastly increased. Brugmann posits five vowels
in long and short form, plus schwa. These are involved in 14 diphthongs, so that the vocalic
system alone includes 25 entities. If we include among the vowels syllabic resonants, the
total is 37. In addition there is a large number of stops, nine fricatives and eight consonantal
resonants. I do not deal here with theoretical views on accent, but I should note that
Brugmann also posited stress and pitch.

When we identify the theoretical bases underlying the massive expansion of the system, we
note that Brugmann’s phonological entities are basically phonetic. The separate entities
representing vocalic and consonantal resonants provide a clear example. Brugmann knew of
their alternation. He cites examples which Sievers had discussed in the article presenting
what came to be known as Sievers’s law. Nonetheless he regarded the alternates as distinct.
Brugmann's basic phonological entity is the 'Sprachlaut', not the phoneme. While ‘Sprach-
laut’ implies a class rather than a simple sound, the class is determined by phonetic criteria
rather than by relationships in a system. For Brugmann, phonological classes are charac-
terized then by their phonetic value not by their interrelationships.

228

In spite of this use of the ‘Sprachlaut’, Brugmann sought system in language. This assump-
tion may be illustrated by the set of vocalic elements, including long vocalic resonants; one
is posited for every characteristic articulatory position, however little evidence we have for
it. Adherence to systematization is almost grotesque in the set of fricatives, which are given
the pattern of the stops; Brugmann finds little evidence for aspirated s and z in his detailed
treatment of the fricatives, yet he posits them. Scrutiny of his phonological treatment leads
one to suggest that he might have done well to avoid the notion of system.

Hirt, Brugmann’s successor in producing a comprehensive Indo-European grammar, follow-
ed similar theoretical procedures (1921—37; 1939: 116—165). He proposes even more
vocalic elements than does Brugmann, including rising as well as falling diphthongs. The
theoretical approach of these two important Indo-Europeanists is based primarily on con-
sideration of phonetic entities which were parallel in their distribution.

3. The classical phonemic approach

Classical phonemics was developed in the twenties and thirties of this century, after Brug-
mann's death and after publication of the phonological volumes of Hirt's Indo-European
grammar. The theoretical effects are already apparent in the great French Indo-Europeanist,
Antoine Meillet, who died in the same year as Hermann Hirt, 1936.

In the 8th edition of Meillet’s Introduction, published under supervision of Benveniste in
1937 with only slight revisions from the 7th edition of 1934, Meillet notes changes in lin-
guistic theory following the publication of Saussure's Cours (1937: XIII—XIV). When,
however, we compare Meillet's phonological system with that of Brugmann, we find only
one major change. Rather than positing three classes of palato-velar obstruents, Meillet
posits two: a labio-velar and a non-labialized palato-velar. This system accords with the two
oppositions found throughout the Indo-European area: in some languages, between labio-
velars and non-labio-velars, in the others, between palatals and velars. In keeping with Saus-
sure’s insistence on the importance of oppositions, only two sets are necessary, as presented
by Meillet (1937: 93—94).

Most subsequent Indo-Europeanists interpret the data as did Meillet. The individualistic
Polish linguist, Jerzy Kuryłowicz, also posited only two sets (1935: 1--26). But these are
palatals and velars. For him the languages with labio-velars represent innovation.

While Meiilet's revised system shows only slight modification by the phonemic principle,
subsequent scholars applied the principle much more thoroughly. Possibly the most signifi-
cant application is Franklin Edgerton's, to the resonants. An introductory article on the
problem in 1934, 'Sievers's law and Indo-European weak grade vocalism', was followed by
an extensive essay in 1943, 'The Indo-European resonants'. In these articles Edgerton
analyzed the Indo-European system rigorously in accordance with the principles of classical
phonemics.

Following Sievers's conclusions based initially on Gothic, Edgerton concluded that con-
sonantal and vocalic forms of the six resonants stand in complementary distribution, as
reflected in Gothic data.

229

Table 3
Gothic data illustrating variation of Gmc i:j:ÿ
lst sg. pres. act. nasja stoja sökja (Edgerton B)
2ndsg. pres. act. nasjis stôjis ^ (sokeis « *iji Edgerton C)
lst sg. pret. act. nasida stauida sökida (Edgerton A)

Relying especially on Vedic data, he assumed for Proto-Indo-European three allophones for
each resonant: A. vocalic, B. consonantal, C. vocalic + consonantal. These are precisely
specified in a series of formulae like the following (1943: 108):

A tit /itti/ (/ * word boundary)
B aya /yaay/
C ätya butktiya àtiya/ tiya

By this view twelve of Brugmann's and Meillet's phonological entities are reduced to six.
The formulation had a powerful appeal. The Indo-European phonemic inventory seemed
more realistic with its smaller numbers. The variants were governed by precise rules. Sub-
sequently this description of variation in the Proto-Indo-European resonants is often
referred to as the Sievers-Edgerton law.

Acceptance of the Sievers-Edgerton law entails further modifications of Brugmann's in-
ventory. It eliminates the diphthongs. If -y- and -w- are variants in well-defined environ-
ments, their combination with preceding e and o cannot be viewed as distinct phonological
entities.

Moreover, the vowel system is reduced. There is no basis for proposing the five short
vowels of Latin for Proto-Indo-European. Only three may be assumed.

Further, the environments for Edgerton's set here labeled C might be explicated. If vowel
length plus consonant is equivalent to consonant plus consonant, the environment might
actually be two preceding consonants or their equivalent. Rather than long vowels as in Go
sokeis, combinations of short vowel plus a consonantal element might be assumed for the
time when the law was in effect. That consonantal element might be equated with Saus-
sure's coefficients, Moller's laryngeals.

These in the meantime had gained greatly in credibility by Kurylowicz's publication of
1927 pointing out that Hittite texts included forms for which Saussure had posited a coef-
ficient. Many of these forms had a consonant at the place of the coefficient, a consonant
indicated by h in our transcribed texts (e.g. Kurylowicz 1935: 73—74). Orthographic evi-
dence had been established for the consonantal elements which Saussure had proposed on
the basis of the method of internal reconstruction.

The Sievers law phenomena in this way added credence to the laryngeal theory, at the same
time providing evidence for different functioning of the resonants from the laryngeals.
Four sub-sets may then be proposed for the Proto-Indo-European phonological system:
Obstruents, Laryngeals, Resonants, Vowels.

230

Table 4
System taking Sievers’s law and the laryngeal theory into consideration
1. Obstruents: p t k kW
(0 d g g”
ph gh gh gwh
$
2. Laryngeals: ?
x h
Y
3. Resonants: m n
wr l y
4. Vowels: i: u:
ee ao o
a:a

Controversies concerning the laryngeal theory are of lesser concern to us here, though vary-
ing views are of interest in illustrating the effect of differing theoretical approaches. Some
Indo-European phonologists continued to treat the laryngeals as abstract entities, e.g. Ben-
veniste and Kurylowicz, who used the symbols: 9; 99 93 (94). Others identified them
phonetically, e.g. Sapir and Sturtevant. Some posited sub-systems of lesser or of greater
numbers, ascending as high as 8. Determination of the number of laryngeals rests in part on
reconstruction of several prehistoric stages (cf. Lehmann 1952: 112—114; Meid 1973:

204—219; Neu 1976: 239—254) or only one, in part on one's theoretical views concerning
system in language. Sapir in his balanced way assumed laryngeals for which he found
reflexes and also parallels in other languages (1938); in this way he introduced typological
considerations, which subsequently became highly prominent in treatments of the early
system. With arguments that are difficult to ignore, he proposed four laryngeals, as given
in Table 4.

Phonologists were aware of problems with this system. The realization of the third row of
obstruents was unclear. As long as the previous system including ph th kh kWh was main-
tained, the symmetry provided cogent support for a balanced set of entities; one could
even assume realization of the sets transliterated with h as voiceless and voiced aspirated
stops. But when the ph series was eliminated, after acceptance of the laryngeal theory, a
system including voiced aspirated stops in contrast with voiceless and voiced stops seemed
highly unusual. Even earlier some linguists, among them Eduard Prokosch, held these
entities to be voiceless fricatives, buttressing this assumption with reflexes like Latin frater
with its f£ (1939: 57—59). No solution was however widely accepted.

A second problem was given less attention at this time, the lack of evidence for 5 noted
already by Schleicher. In general, Indo-Europeanists pointed to this oddity but did little
more about it.

A further concern arose from scrutiny of the Vedic texts for treatment of the resonants.
The Vedas contain many examples of the functioning of Sievers's law, but there is also con-
tradictory evidence. For Edgerton, the contradictory evidence is found in late texts, pro-
duced after a time when the law was in force. The aberrant patterns seemed to him parallel
to umlaut introduced in many forms which lacked the phonological structure triggering

231

it, such as German Generäle ‘generals’ with umlaut in a word brought into German after
the period of the umlaut changes and Löcher ‘holes’ with umlaut in a pattern that should
have umlaut of u rather than o, and many more, restructured by the process traditionally
referred to as analogy. Other scholars took a different view, regarding the Vedic situation
as more accurate evidence than the theoretical assumptions of Sievers, Edgerton, and those
who accept Sievers’s law.

The conflict between the two views intensified, leading to revisions in the theoretical ap-
proach of linguists such as Kurylowicz. By this approach which Kurylowicz allowed in his
later years, many phonological alternations in morphological sets were to be explained
Without assuming origin in a phonological change, but rather through morphological
structuring alone. As a further example we may cite Brugmann’s law, a proposal to account
for a rather than short a as the reflex in some Sanskrit forms for Proto-Indo-European o.
The perfect provides examples illustrating the difficulty. This commonly has a short vowel
in the first singular, as in tatana (also however tatana) ‘I stretched', but long vowel in the
third singular, as in tatana ‘he stretched’. In its strict form Brugmann’s law proposes change
of short PIE o to Skt. a before resonants in open syllables. Before the laryngeal theory
was proposed, the contrast between the first and third singular stem vowels was unex-
plained. After the Hittite h-conjugation came to be known, with -hi in the first singular,
4 in the third, it became clear that the first singular had o in a checked rather than in an
open syllable, and accordingly was not subject to Brugmann's law.

Like the Sievers's law phenomena, however, the expected reflexes of Brugmann's law fre-
quently are not found. Moreover, long 2 for PIE o is attested in some forms before ob-
struents. By the classical theory of historical linguistics, the aberrant forms are accounted
for by analogy; at one time in early Indic, short o in well-defined environments became
long 4. By the contrasting view, there never was a phonological change; all the long 4 were
morphologically triggered.

The expanded role of morphology may seem based on the generative view of systematic
as opposed to classical phonemics. The generativists however did not concern themselves
to any extent with Proto-Indo-European phonology; it is difficult to cite any work by lin-
guists practicing the generative transformational approach which contributed to our under-
standing of Proto-Indo-European segmental phonology. It is possible that linguists outside
the central generative transformational sphere, such as Kurylowicz, took over some notions
of systematic phonemics. In any event, the confrontation between the two approaches may
have led to acceptance of greater flexibility in applying classical historical theory.

The flexible view may be summarized as follows:

(1) Phonological contrasts may be introduced into a linguistic system. The contrast may
arise variously: a) by a sound change in the classical sense, as when American English
medial /t/ in specific environments falls together with /d/ for many speakers, e.g. bitter :
bidder, beatle : beadle; b) by introduction of a contrast through borrowing, whether from
another dialect or another language, e.g. English /Z/ as in beige.

(2) The change may be established and extended to further speakers, or it may be elimi-
nated.

(3) If it is established and extended, it may furnish the basis of expressing specific mean-
ings, whether grammatical or lexical. Such extension for morphological purposes has not

232

happened for English /d/ as opposed to /t/. We could however develop a further means of
distinguishing between present and past from the phonological contrast t : dad in such
pairs as fit : fitted, seat : seated. Such a contrast would be useful in overcoming the ambi-
guity in current English verbs like beat; it may well be that such verbs are too few to
warrant a distinction. If however speakers perceive a need for a specific morphological
contrast, they may adopt one from such phonological contrasts in a language.

Umlaut in the late Germanic languages provides an example we have already drawn on.
The source of the change is assimilation of stem vowels to high vowels in the remainder of
the word, either later i or u. The umlaut change became phonemic through a further
change — weakening and often loss of the triggering vowels; by that loss a means for
distinguishing some plurals from singulars was also forfeited. Umlaut of stem vowels
provided a means for a new plural marker. Its morphological application in English plurals
like men : man, geese : goose, mice : mouse is relatively limited; but Standard German
provides many nominal examples, and also examples in the verb, like that between the
third singular present of geben : gibt, and the second plural gebt. A clear statement of this
classical view was given by Weinreich, Labov and Herzog in 1968, especially the processes
of embedding and extending the results of a change.

As a result of scrutiny of specific problems in accordance with the phonemic approach, the
theoretical position of Indo-European phonologists increased in explicitness as well as in
flexibility. After Schleicher the great advance came in careful phonetic analysis. The
phonemic approach yielded credible systems from the large numbers of phonetic elements
posited by Brugmann, Hirt and their contemporaries. It was however applied almost ex-
clusively to data in the Indo-European language family. Subsequently phonological theory
has been reviewed in accordance with the findings of typology.

4. Typological contributions

By a typological approach one compares entities for specific properties, attempting in this
way better to understand those entities within a given system and given range of functions.
One could, for example, attempt a typology of vehicles, classifying them for physical or
surface characteristics. A deeper classification would take function into consideration.
Additional scrutiny might involve attention to a social group, which like the early Indo-
European speakers made use of wagons and chariots in contrast with many of their con-
temporaries. Typology then does not confine itself to external form.

Typology has been pursued since the early days of modern linguistics, paralleling the pre-
dominant concerns of the discipline though never as prominent as historical or purely
descriptive study. Beside Grimm we can place von Humboldt, who was followed by the
eminent scholars Steinthal, Finck, W. Schmidt and Sapir. However brilliant, the typologists
occupy a secondary position beside linguists like Grimm, Bopp, Schleicher, the neogram-
marians and their students, such as Leonard Bloomfield and current scholars who disregard
the approach. Roman Jakobson has stated for example that the typologists were “distrusted
by the neogrammarians" (1962: 524).

Jakobson made this statement at the International Congress of Linguists at Oslo in 1957.
Possibly because of the forum provided there his remarks influenced subsequent work on

233

Indo-European phonology, while Prokosch and others with similar views on the Proto-
Indo-European obstruent system were largely disregarded. One of Jakobson’s statements
is on noted: “To my knowledge, no language adds to the pair /t/ — /d/ a voiced
aspirate /d'!/ without having its voiceless counterpart /thy , while /t/, /d/, and Ahi frequent-
ly occur without the comparatively rare Jdhj ... ; therefore theories operating with the
three phonemes /t/ — /d/ — /dh/ in Proto-Indo-European must reconsider the question of
their phonemic essence" (1962: 528).

It was not this feature, however, that aroused new consideration of the Indo-European
obstruent system, but rather the absence of /b/, which Schleicher had already noted. In the
meantime typologists had found such a hole in the set of labial obstruents when the series
involved was glottalized. Yet the absent glottalized labial was not voiced, but rather voice-
less. This observation led to the glottalic theory for pre-Indo-European.

Much like the law of palatals in the 1870s, the glottalic theory was proposed simultaneous-
ly by different scholars, Paul Hopper in this country and Tamas Gamkrelidze with Vyaches-
lav Ivanov in the USSR. Their initial articles were published in 1973, and have been fol-
lowed up by further publications. The glottalic theory assumes that b was missing in Proto-
Indo-European because the earlier system included glottalic voiceless stops rather than a
series of voiced stops. (To maintain clarity of presentation, the old labels of Grimm have
been taken over for the three kinds of obstruents, as in a recent paper by Hopper (1982).)
The sub-sets may be illustrated by the dentals.

Tenues Mediae Mediae Aspiratae

Pre-Indo-European t t d
Proto-Indo-European t, th d dh
Taihun (Proto-Germanic) — p t d
Decem (Sanskrit) t, th d dh

In this way the glottalic theory has suggested a solution for the ancient problem of missing
b.

The theoretical basis for the glottalic theory differs markedly from those on which the
modifications were proposed in the neogrammarian periods and the period of structuralism.
Each of these theoretical approaches made use of the primary methods applies in historical
linguistics, methods which Hoenigswald has acutely scrutinized and elaborated (1960,
1973). Both the comparative method and the method of internal construction require data
in the languages to which they are applied. Conjectures regarding the obstruent system, on
the other hand, are derived in the absence of such data; reconstruction of the “mediae
aspiratae” to be sure is suggested on the basis of reflexes in some dialects though as past
consideration has shown the data leave broad rein to the imagination; data for reconstruct-
ing glottalics are absent, permitting inferences based only on generalizations drawn from
scrutiny of phonological systems in language.

Not surprisingly the glottalic theory has encountered considerable scepticism, as may be
concluded from the successive publications of the principal proponents of the theory.
Fortune may bring support like that of Hittite for Saussure’s coefficients. The chances are
greater for evidence from internal features of the known dialects which may provide such
support through clues interpreted with improved understanding of language based on
increasing scrutiny of phonological characteristics in typological study. Even then strong

234

support will be needed to offset the scepticism resulting from the greater complexity of
historical developments required by the glottalic theory than from a system of Proto-Indo-
European obstruents like that proposed by Prokosch. Formerly Germanic and Armenian
were assumed to have undergone far-reaching obstruent shifts, as formulated in Grimm’s
Law and the rules for the Armenian shift. Now such shifts must be assumed for all sub-
groups of the family. The older view accordingly represented far greater economy.

Further, proponents of the glottalic theory can provide no support other than that from
typology. They are accordingly in the position of Saussure in 1878, whose coefficients
were proposed purely on a theoretical basis. But for the glottalic proponents there has not
been a Möller to identify their newly proposed consonants with possible cognates in an-
other set of languages like the Semitic.

Summing up, we note three major periods and kinds of innovative theory.

(1) After Schleicher's proposal of a phonological system for Proto-Indo-European, great
elaboration was provided by the increased understanding in phonetics. The phonetic hand-
books of Sievers, Jespersen and Sweet have still scarely been surpassed for articulatory
phonetics. Illustrations of their mastery are available in contemporary articles on Indo-
European phenomena.

(2) Development of phonemic theory contributed recognition of interrelationships between
the many entities posited by the phonetic approach. It also led to a compact system com-
posed of sub-systems. This system was examined for morphophonemic relationships. As
sketched above, the morphophonemic relationships treated in Sievers's law helped support
the laryngeal theory. This theory gained further credence from its elucidation of other
phonological cruxes, such as the Brugmann's law phenomena, and from its clarification of
relationships between morphological sets, such as the fifth, seventh and ninth verb classes
of Sanskrit. Yet the massive evidence from morphology led some linguists to locate the
basis of phonological cruxes in the morphological rather than the phonological sphere
(Kurylowicz 1977: 6—52, and elsewhere; see also Seebold 1972). The clash in theoretical
positions has by no means been resolved, though both sides proposed a relatively similar
phonemic system for Proto-Indo-European.

(3) To this phonemic system with its detailed phonetic analysis the results of typological
theory are being applied, especially to earlier stages of the system. Assuming continued
scrutiny comparable to that in former innovative concern, we may expect continued atten-
tion to typology; this may yield improved understanding of phonological cruxes, such as
those in the suprasegmental system, and as a result gain greater credence for the glottalic
theory.

These brief considerations of theoretical views may illustrate how differences in theoretical
position affect the conclusions based on data, even though the data themselves are scarely
modified. Such effects may be most noticeable in dealing with reconstructed languages
though they are not absent from treatments of languages or other social systems directly
available to us.

235
References

Bechtel, Fritz, 1892. Die Hauptprobleme der indogermanischen Lautlehre seit Schleicher (Göttingen:
Vandenhoeck & Ruprecht).

Brugmann, Karl, 1904. Kurze vergleichende Grammatik der indogermanischen Sprachen (Straßburg:
Trübner).

Edgerton, Franklin, 1934. “Sievers’s Law and Indo-European Weak-grade Vocalism", Language 10:
235-653.

Edgerton, Franklin, 1943. “The Indo-European Semivowels”, Language 19: 83—124.

Gamkrelidze, Tamas/Vyacheslav Ivanov, 1973. “Sprachtypologie und die Rekonstruktion der gemein-
indogermanischen Verschlüsse. Vorläufiger Bericht”, Phonetica 27: 150—156.

Hirt, Hermann, 1921-1937. Indogermanische Grammatik 1--7 (Heidelberg: Winter).

Hirt, Hermann, 1939. Die Hauptprobleme der indogermanischen Sprachwissenschaft, edited by Helmut
Arntz (Halle: Niemeyer).

Hoenigswald, Henry M., 1960. Language Change and Linguistic Reconstruction (Chicago: University of
Chicago Press).

Hoenigswald, Henry M., 1973. Studies in Formal Historical Linguistics (Dordrecht: Reidel).

Hopper, Paul J., 1973. “Glottalized and Murmured Occlusives in Indo-European", Glossa 7: 141—166.

Hopper, Paul J., 1981. ““Decem” and “Taihun” Languages: An Indo-European Isogloss”, in: Bono
homini donum: Essays in Historical Linguistics in Memory of J. Alexander Kerns, edited by Yoél L.
Arbeitman & Allan R. Bomhard (Amsterdam: Benjamins), 133—142.

Jakobson, Roman, 1957. “Typology and Reconstruction”, Report of the First Plenary Session of the
Eighth International Congress of Linguists, Oslo, 5 August 1957.

Jakobson, Roman, 1971. Selected Writings 2: Words and Language (The Hague: Mouton).

Kurylowicz, Jerzy, 1927. **o indo-europeen et A hittite", in: Symbolae Grammaticae in honorem Ioannis
Rozwadowski, Vol. 1 (Kracöw: Uniwersytet Jagiellonski), 95-104.

Kuryłowicz, Jerzy, 1935. Etudes Indoeuropéennes (Kraków: Polska Akademja Umiejętności).

Kuryłowicz, Jerzy, 1977. Problèmes de Linguistique Indo-Européenne (Warsaw: Polska Akademia Nauk).

Lehmann, Winfred P., 1952. Proto-Indo-European Phonology (Austin: University of Texas Press).

Meid, Wolfgang, 1974. “Räumliche und zeitliche Gliederung des Indogermanischen”, in: Flexion und
Wortbildung, edited by Helmut Rix (Wiesbaden: Reichert) 204—219.

Meillet, Antoine, 1937. Introduction à l'étude comparative des langues indo-européennes, 8th edition
(Paris: Hachette).

Möller, Hermann, 1879. Englische Studien 3: 157n.

Neu, Erich, 1976. “Zur Rekonstruktion des indogermanischen Verbalsystems”, in: Studies in Greek,
Italic, and Indo-European Linguistics Offered to Leonard R. Palmer, edited by Anna Morpurgo Davies
and Wolfgang Meid (Innsbruck: Institut für Sprachwissenschaft), 239—254.

Prokosch, Eduard, 1939. A Comparative Germanic Grammar (Philadelphia: University of Pennsylvania
Press).

Sapir, Edward, 1938. ‘‘Glottalized continuants in Navaho, Nootka, and Kwakiutl”’, Language 14: 248—
274.

Sapir, Edward, 1939. “From Sapir’s desk", Language 15: 178—182.

Saussure, Ferdinand de, 1879, 1887. Mémoire sur le système primitif des voyelles dans les langues indo-
européennes (Paris: Vieweg).

Saussure, Ferdinand de, 1949. Cours de linguistique générale (1916, 1st edition) (Paris: Payot).

Schleicher, August, 1871. Compendium der vergleichenden Grammatik der indogermanischen Sprachen
3d edition (Weimar: Böhlau).

Seebold, Elmar, 1972. Das System der indogermanischen Halbvokale (Heidelberg: Winter).

Weinreich, Uriel/William Labov/Marvin I. Herzog, 1968. “Empirical Foundations for a Theory of Lan-
guage Change”, in: Directions for Historical Linguistics, edited by Winfred P. Lehmann and Yakov
Malkiel (Austin: University of Texas Press), 95—195.

L’articolo indeterminativo
(note per la storia della grammatica italiana)

Giulio C. Lepschy (University of Reading)

A prima vista, l’esistenza e il valore dell’articolo indeterminativo (un), a fianco di quello
determinativo (il), possono apparire ovvi e non problematici, e l'impressione sembra con-
fermata dalle grammatiche scolastiche, e da quelle correnti per lo studio delle lingue
straniere: il corrisponde al francese /e, all’inglese tre, al tedesco der, ecc., e un al francese
un , all'inglese 2, al tedesco ein, ecc.

Ma basta un momento di riflessione per accorgersi che la faccenda non é cosi semplice. T
due articoli non sembrano opporsi l'uno all'altro significativamente, costituendo una na-
turale classe paradigmatica. Nella maggior parte dei casi non si ha scelta, ma è la struttura
dell'informazione nel discorso che impone l'uso dell'indeterminativo per l'introduzione di
un elemento nuovo, e del determinativo per uno già nominato (come indica l'univocità
delle scelte nelle traduzioni da lingue senza articoli a lingue con articoli: cfr. Lepschy
(1980)). Quanto alle implicazioni logiche, basta scorrere un manuale (per es. Reichenbach
(1947)) per vedere che i problemi posti dall'indeterminativo (nei suoi usi, specifico, non
specifico, e generico, e per l'inclusione e l'appartenenza a una classe) sono molto diversi da
quelli delle descrizioni definite che Russell associa all'uso del determinativo (per una
discussione in ambito generativistico cfr. Hawkins (1978); per una sintesi di linguistica
generale cfr. Krámsky (1972); e per l'italiano cfr. Renzi (1976b)).

L'indeterminativo presenta delle difficoltà specifiche. Un italiano che cominci a studiare
l'inglese si troverà incerto se dire, in corrispondenza di un etto, un chilo, ecc., an ounce, a
pound, oppure one ounce, one pound, ecc., e si renderà conto che la sua incertezza non
riguarda soltanto l'uso inglese (se entrambe le forme, con a e con one, siano possibili, e, se
lo sono, come differiscano), ma anche il valore della forma italiana: si tratta dell'articolo
indeterminativo, o del numerale uno? E, di fronte alla difficoltà di dare una risposta, ci si
può chiedere se la distinzione sia necessaria, o, in altri termini, se l’indeterminativo di fatto
esista, o se non si possa farne a meno, e interpretarlo come un numerale. Di fronte a vorrei
una mela, ci si può chiedere a che cosa si opponga questo una: il senso (a) di numerale si
oppone a due, tre, ecc. come indeterminativo si oppone forse a questa, quella, o altri
dimostrativi, almeno nel suo senso (b) di ‘una qualsiasi’; nel senso (c) di ‘una certa’ si
opporrebbe invece al senso (b). D'altra parte, il tipo di opposizione è diverso: mentre
riferendosi al senso (b) sarebbe possibile rispondere: ‘ecco, ti darò questa’, riferendosi al
senso (a) sarebbe strano rispondere: ‘ecco, te ne darò due’. Per il francese l’interessante
grammatica di Martinet osserva:

238

“L’article un [...] est homonyme du cardinal un [...] Il arrive fréquemment qu'on ne
sache auquel on a affaire, ou méme que les deux valeurs soient combinées et présentes dans
l'énoncé" (1979: 43).

Perfino in una lingua come l'inglese, in cui l'articolo indeterminativo (a, art) ha una forma
diversa dal numerale (one), é stato proposto, nell'ambito della grammatica generativa, di
vedere nel primo una realizzazione non accentata del secondo (Perlmutter (1970); e cfr. per
l'italiano le obiezioni di De Boer (1972) e di Renzi (1976a)). Come per molte altre ipotesi
generativistiche à interessante notare che qui abbiamo un'interpretazione che da un lato
ricapitola nella derivazione sincronica uno sviluppo diacronico, e dall'altro riecheggia idee
di grammatici del passato. Questi infatti hanno esitato a lungo, prima di appaiare l'articolo
indeterminativo a quello determinativo, ed è appunto sulle loro discussioni che intendo
soffermarmi in questo saggio.

Come è noto, presso i grammatici greci arthron poteva comprendere, oltre all’articolo,
anche pronomi di vario tipo (insieme ad altre parti del discorso), e i grammatici latini,
trasferendo nozioni e termini greci al latino, che per di più non aveva l'articolo, non intro-
ducono chiarezza in questo settore. Le prime descrizioni grammaticali dei volgari europei
moderni, impostate sulle grammatiche latine, e applicate a lingue che, a differenza del
latino, usano gli articoli, rivelano considerevoli oscillazioni. Per il francese la storia del-
l’articolo nella tradizione grammaticale è stata tracciata da Yvon (1955—1956); e per
l'inglese la classificazione dell'articolo fra le parti del discorso viene tabulata nella mono-
grafia di Ian Michael (1970: 532). L’articolo indeterminativo viene a volte associato a
quello determinativo, ma stenta a farsi strada. Nella tradizione inglese i termini ‘finite’ e
‘infinite’ si trovano in Ben Jonson (1640), e ‘definite’ e ‘indefinite’ in J. Howell (1662).
In quella francese un viene considerato un articolo ‘improprio’, e i termini ‘infini’, ‘in-
défini” vengono per lo più riferiti alle forme invariabili, cioè alle preposizioni. Bisogna
arrivare alla Grammaire générale di Port-Royal per trovare una descrizione che a noi appare
singolarmente moderna, in cui si oppone l’articolo ‘défini’ (Ze, la) a quello ‘indéfini’ (un,
une). Ma le principali grammatiche successive, anche per influenza di Du Marsais e del-
l’Enciclopedia, accettano solo l’articolo determinativo, come, tanto per dare un esempio,
nella Grammaire des grammaires di P. Girault-Duvivier (1812), e, fra la grammatiche
pedagogiche, un quella diffusissima di F.J. Noél e Ch. P. Chapsal (1823, ristampata fino al
1888), che non parla affatto dell’indeterminativo.

Per la tradizione italiana, Kukenheim (1932: 126) ha osservato che, nonostante l’acume
rivelato dai nostri grammatici, si è tardato molto a rendersi conto della funzione particolare
dell’articolo indeterminativo. Va notato peraltro, insieme al ritardo, un anticipo, se è vero
che la prima discussione importante sull’uso di un, è quella di Salviati (1586).

Senza pretendere di fare un esame sistematico, ho dato una rapida scorsa a varie gram-
matiche che mi erano facilmente accessibili (per un regesto, generalmente attento anche
alla questione dell'articolo, si veda la vecchia ma insostituita Storia del Trabalza (1908).
Fin dall'inizio, si trova, nella presentazione dell'articolo, una serie di preoccupazioni
diverse:

(a) l'articolo come particella che serve ad illustrare il genere, il numero e il caso della
parola, un po' come in Prisciano la declinazione era accompagnata dalle forme di hic (che
del resto era chiamato ‘articulus’, o ‘pronomen articulare’): hoc mare, huius maris, huic
mari, ab hoc mari; e cosi el cielo, del cielo, al cielo, dal cielo. Questo porta ad associare

239

l’articolo alla preposizione, e ad esaminare l’uso delle preposizioni articolate. Naturalmente
a noi questa impostazione sembra trascurare la disparità fra il latino, in cui le desinenze dei
nomi indicavano simultaneamente genere, numero e caso, e l’italiano in cui esse indicano
genere e numero, ma non caso.

(b) Vengono precisate le forme diverse che l’articolo (con le sue variazioni di genere e di
numero) può avere (per es., il o e/), anche a seconda delle parole a cui si accompagna (per
es., il o lo);

(c) vengono illustrate le regole per la presenza o l'assenza dell'articolo, spesso con grande
penetrazione, e considerando anche complesse questioni sintattiche (si pensi ai casi come
le immagini della cera, una immagine di cera esaminati dal Bembo);

(d) oltre che alla preposizione, l'articolo viene spesso associato al pronome, in particolare
per i casi, che a noi appaiono di omofonia, di i], lo, la, ecc. come complementi oggetti, che
vengono di solito interpretati come esempi di articoli usati in funzione di pronome;

(e) a volte vengono presentate norme che a noi non sono immediatamente ovvie, e che
meriterebbero ancora di essere chiarite (per es., le osservazioni del Fortunio (1516: XI!)
sulle forme che vanno con gli aggettivi piuttosto che coi sostantivi; o quelle del Bembo
(1525: 199 dell’ed. Dionisotti) sulla e che si introdurrebbe al posto della vocale iniziale
di parola e di quella finale dell’articolo).

Ma si tratta, in ogni caso, dell’articolo determinativo. Non ho trovato menzione di quello
che noi chiamiamo indeterminativo, non solo nelle prime grammatiche, ma anche in molte
delle trattazioni più mature: per es., nell’opuscolo di L.B. Alberti (della metà del Quattro-
cento), nelle Regole del Fortunio (1516), nelle Prose del Bembo (1525), nelle Tre fontane
del Liburnio (1526), che elencano gli articoli fra i pronomi, nella Grammatichetta del
Trissino (1529), nella Grammatica volgare dell’Acharisio (1537), in quella di Tizzone
Gaetano (1539), nelle Regole di lacomo Gabriele (1545), nella Grammaire italienne di J.P.
de Mesmes (1548), nei Fondamenti di R. Corso (1549), nelle Osservationi del Dolce (1550),
in De la lingua di P.F. Giambullari (1551), nella Giunta del Castelvetro (1563), nell'7talian
Grammar che H. Granthan (1575) basava sui Praecepta (1567) di Scipione Lentulo.

Ma anche dopo il Salviati continua a colpirci l’assenza di una trattazione adeguata di un,
anche dove più ci si aspetterebbe di trovarla, come nella Nouvelle méthode italiana di Port-
Royal (1660). Il Mambelli (1644) parla di uno come “pronome universale indeterminato”
(cito da un’edizione del 1709, p. 392), e M. Kramer (cito dalla quindicesima edizione, del
1763, della sua Vollkommene toscanisch- und romanische italiänische grammatica) distin-
gue “Articul oder Bestimmwoertlein” (lo, l2) (p. 38), "Articulo indefinito” (come nella
declinazione Dio, di Dio, a Dio, Dio, da Dio) (p. 48), e "Pronom. Indefinitum" (uno), che
chiama anche "Articulus unitatis" (p. 377). Ma in generale l'indeterminativo viene ignora-
to, come per esempio nella Grammar of the Italian Tongue di John Henley (1719), nelle
Regole di G. Gigli (1721), in Della lingua nobile d’Italia di N. Amenta (1724), nel Traité
dell’Antonini (1726) dove si chiamano articoli indefiniti, secondo la tradizione francese, le
preposizioni o l’assenza dell’articolo. Il Baretti nella sua grammatica italiana (1760) non
parla dell'indefinito, e nella grammatica inglese osserva che l'articolo inglese @ “ha un
significato indefinito, e vuol dire, one, cioè uno o una, relativamente a more, più”. Così
non ne parla il Maître italien (1787) del Veneroni, né il Dictionary of the Peculiarities of
the Italian Language (1820) del Santagnello, che per l’inglese a, an dà “uno, un, una,
declined with the indefinite article di, a, da" (p. 39); né gli Avvertimenti gramaticali (1826)

240

di Sforza Pallavicino; né la Grammaire italienne del Vergani, che ho consultato in un'edi-
zione londinese del 1836. E" vero che la Grammatica del Soave (1771, ma cito da un'edi-
zione del 1815) distingue fra articoli determinati" e “indeterminati”: si usa uno, una
“quando fra gli oggetti compresi sotto ad un nome universale, non si pensa che ad indicarne
uno qualunque” (p. 8). Ma solo intorno alla metà del secolo si stabilizza la situazione che a
noi appare normale: la grande Grammatik der italiänischen Sprache (1844) di L.G. Blanc
dà i due articoli, “bestimmte” (o come meglio si dovrebbe dire “bestimmende”), e “unbe-
stimmte”, corrispondenti al tedesco der e ein, e osserva che “Letzterer, der unbestimmte,
wird durch das Zahlwort uno, a vertreten: /'uomo (der Mensch), un uomo (ein Mensch)"
(p. 168). Una nota avverte che il milanese ha una forma particolare per l'articolo (on, ona),
distinta da quella del numerale (vun, vuna) (cfr. p. 643). Il Gherardini, nella Appendice
alle grammatiche italiane (1843) nota che “Li Articoli sono dunque determinativi come Il,
Lo, La, ec. Nondimeno si sogliono chiamare Articoli anche le voci Uno e Una, quantunque
non valgano a determinar precisamente veruna cosa; onde son detti Articoli indetermina-
tivi” (pp. 57--58). Cosi i due articoli si trovano in uno dei manuali più fortunati, la /talian
Conversation-Grammar (1858) di Charles Marquard Sauer (dove si usano i nomi ‘articolo
definito" e “indefinito”, in corrispondenza a quelli inglesi “definite” e “indefinite”); nella
Grammatica italiana di R. Fornaciari (1879) in cui si usano i termini ‘articolo determina-
to” e “indeterminato” (e si collegano gli articoli ai pronomi, dimostrativo e numerale
rispettivamente); con i termini “determinativo” e ‘indeterminativo” nella Grammatica
(“Approvata come libro di testo dal ministero della Pubblica Istruzione”) di Cesare Maria-
ni, uscita all’inizio degli anni Ottanta (cito da un’edizione del 1904); e con i termini “deter-
minativo o determinato”, “indeterminativo o indeterminato” nella Grammatica scolastica

(1896) di Ettore Piazza (che cito da un’edizione del 1905).

Come ho detto sopra, la prima trattazione importante a me nota di quello che oggi chia-
miamo l'articolo indeterminativo, à dovuta al Salviati, nel secondo volume degli Avverti-
menti (1586). Qui l'indeterminativo non viene accoppiato al determinativo, ma viene messo
in una categoria a parte, chiamata “accompagnanome”, di cui si indicano le somiglianze e le
differenze rispetto all’articolo, con un acume e una precisione da cui possiamo imparare
anche oggi. Vale la pena di riportare il passo, che non richiede chiarimenti:

“Dal Nome, nell’opera del sentimento, tuttochè nome sia anch'ella, è forse da distinguere
una certa parte del favellare, che accompagnanome in questi libri ci piace di nominarla:
posciachè proprio titolo non l’è ancora, che noi sappiamo, stato dato nel volgar nostro: ne
dal Latino, à dal Greco, il possiamo torre in prestanza, che cotal parte non usarono in lor
sermone, ne conoscerla, non che nomarla, non poteron per conseguente. Ed è questa, che
noi diciamo, la voce uno, ó una, quando non come numerale, ma per una cotale accom-
pagnatura si mette davanti a nome, che si ponga nel minor numero: che dirado non v'aven-
do l'articolo, senza essa lo troverrai: ma con esso articolo non vi puo mai aver luogo [. . .]
alcuna forza porta nel sentimento, a quella dell'Articolo non intutto dissomigliante. per-
ciocché ristrigne anch'ella al Nome, come l'Articolo, e gli determina il suo valore: ma in cio
sono diversi, che l'Accompagnanome gliele ristrigne, e gliele determina solamente: là dove
l'Articolo, e gliele ristrigne, e gliele determina, e oltr'a questo gliele specifica, e, come da
noi conosciuto, il ci pone avanti nel favellare" (pp. 51--52).

Segue un esempio dal Decameron (II, 8, 87): con a guisa di ragazzo “quasi l’idea del ragaz-
zo esprimerremmo in confuso”; a guisa d'un ragazzo "mostra, che chi lo nomina abbia
nell'animo una sembianza d’un particolar ragazzo, tuttavia che l’uditore non sappia egli gia

241

quale"; mentre a guisa del ragazzo “n’avrebbe disegnato uno, non solamente da chi lo
nomina, ma conosciuto ancora spezialmente da chi sente nomarlo” (p. 52). Ci limiteremo
a notare l’anticipazione della concezione moderna di un per il nuovo, e di i] per il noto (e
ricordiamo a questo proposito le distinzioni che il Castelvetro nella Giunta (1563, p. 11V)
faceva, a proposito di quello e di il, fra tre valori: preterito, futuro, e presente, che possono
corrispondere a quelli che oggi si direbbero anaforico, cataforico, e deittico).

Questa categoria di accompagnanome viene usata nel Vocabolario della Crusca (1612), per
indicare uno dei valori di uno; e viene ripresa dal Buommattei, Della lingua toscana (1643),
fra i “ripieni”: “Ripieno è una particella, non necessaria alla tela grammaticale: ma serve
all’ornamento della frase, per proprietà di linguaggio” (p.368). Qui si perdono le utili
precisazioni del Salviati, ma si aggiunge che questi uno, una, non sono numerali corrispon-
denti al latino unus, una, come si vede provando a tradurre in latino gli esempi italiani in
cui uno è “ripieno”. Le stesse nozioni vengono riprese nelle Regole ed osservazioni (1745)
del Corticelli. Deve essere per influenza della tradizione francese a cui abbiamo accennato
(con l’impostazione di Port-Royal, che accosta i due articoli, e il rifiuto dell’indetermina-
tivo ad opera di Du Marsais) che troviamo considerazioni come quelle di R. Zotti nella sua
Grammaire italienne et francaise (cito dall'edizione del 1836): A l'égard des mots un, une
(uno, una), c'est confondre toutes les notions que de les regarder comme des articles, puis-
que, s'ils en sont, on sera forcé de donner ce nom à presque tous les autres pronoms, tels
que tutto, tout; ogni, chaque" ecc. (pp. 42—43). E, ancora nel 1841, troviamo nella ano-
nima Grammatica ideologica o sia le leggi comuni d'ogni parlare dedotte da quelle del
pensare (su cui cfr. Trabalza (1908, pp. 457—462)), che al determinativo viene rifiutato il
suo valore generico, e dell'indeterminativo viene negata la possibilità:

“Gli articoli pertanto od aggettivi indicativi o determinativi devono tutti afficere la sola
estensione de” nomi generali, ed essere determinativi di questa, sicchè sarebbe un con-
trasenso il chiamarne alcuni indeterminati, salvo in quantochè gli uni determinano men
precisamente che gli altri, onde a petto di questi pajono indeterminati" (p. 201). “L’artico-
lo i, lo, la non parmi punto determinativo d’estensione, se non quando adoprasi in senso
equivalente all’i/le, illa, illud, latino, ond'ei provenne, quando varrà cioè quello, quella, o
la cosa già discorsa, già intesa, indicata o nota comunque |. . .] Non è pertanto vero, che
l'il, lo, la serva anche per far pigliare il nome comune in tutta quanta la sua estensione (i
cani, quasi valga, tutti i cani)" (p. 203). E per l’indeterminativo: “A che pro contendere se
uno possa essere articolo? certamente che uno conformandosi al suo sostantivo notane il
genere e numero e colpisce la sola estensione, ma non la comprensione de' nomi [. . .] Per-
tanto ha sempre in realtà il valore numerale, ma senza contraddistinguere, anzi lasciando in
incerto a quale individuo fra i molti d'una classe s'applichi, perché al proposito del discorso
ciò non importi [. . .] Siavi pur differenza tra l'uno usato rigorosamente come numerativo
(Quanti cavalli hai? ne ho uno), e l'uno usato indeterminatamente (un cavallo mi calpestò;
un uomo mi feri: dove l'uno serve soltanto a notare che l'estensione del suo appellativo é
ridotta ad un solo individuo, ma ad uno qualunque del vasto genere): ma ció non fa che
non sia sempre numerativo e pel numero singolare; é la voce stessa usata sotto due punti di
vista o rapporti, accessori differenti, ma nella sostanza convengono” (pp. 204—206). A
quest'epoca, d'altra parte, la battaglia contro l'indeterminativo era ormai perduta. La
battaglia, ma, come si é visto all'inizio, non la guerra.

242
Bibliografia
Indico nella bibliografia solo le opere pertinenti per l'indeterminativo, e poche altre da cui ho citato.

Bembo, P., 1525. Prose della volgar lingua, in: Prose e rime, a cura di C. Dionisotti (seconda edizione
accresciuta 1966) (Torino: UTET).

Blanc, L.G., 1844. Grammatik der italiänischen Sprache (Halle: Schwetschke).

Buommattei, B., 1643. Della lingua toscana (Firenze: Zanobi Pignoni).

Castelvetro, L., 1563. Giunta fatta al ragionamento degli articoli et de verbi di Messer Pietro Bembo
(Modena: C. Gadaldino).

De Boer, M.G., 1972. "Il concetto di articolo con speciale riguardo all'italiano", SJLTA 1: 511—536.

Du Marsais, C.C., 1769. Logique et principes de grammaire (Paris: Briasson).

Fortunio, G.F., 1516. Regole grammaticali della volgar lingua (Ancona: Bernardin Vercellese).

Gherardini, G., 1843. Appendice alle grammatiche italiane (Milano: Bianchi).

Grammatica ideologica, 1841. Grammatica ideologica o sia le leggi comuni d'ogni parlare dedotte da
quelle del pensare (Milano: Chiusi).

Hawkins, J., 1978. Definiteness and Indefiniteness: a Study in Reference and Grammaticality Prediction
(London: Croom Helm).

Kramer, M., 1763. Vollkommene toscanisch- und romanische italiánische grammatica (Nürnberg:
Schwarzkopf).

Krámsky, J., 1972. The Article and the Concept of Definiteness in Language (The Hague: Mouton).

Kukenheim, L., 1932. Contributions à l'histoire de la grammaire italienne, espagnole et francaise à
l'époque de la renaissance (Amsterdam: Noord-hollandsche uitgevers-maatschappij).

Lancelot, C./ Arnauld, A., 1660. Grammaire générale et raisonnée (Paris: Le Petit).

Lepschy, G., 1980. L'uso dell'articolo: confronti interlinguistici, in: Recherches de linguistique. Hom-
mage à Maurice Leroy (Bruxelles: Editions de l’Université), 119-124 (anche in Lepschy, G. (1981)
Mutamenti di prospettiva nella linguistica (Bologna: il Mulino), 167-172).

Mambelli, M., 1709, Osservazioni della lingua italiana raccolte dal Cinonio (Ferrara: Pomatelli).

Martinet, A., 1979. Grammaire fonctionnelle du francais (Paris: Crédif/Didier).

Michael, I., 1970. English Grammatical Categories and the Tradition to 1800 (London: Cambridge Uni-
versity Press).

Perlmutter, D.M., 1970. On the Article in English, in: Progress in Linguistics. A Collection of Papers,
edited by Bierwisch, M., Heidolph, K.E. (The Hague: Mouton), 233—248.

Renzi, L., 1976a. “Uno: numero e articolo", RGG 1: 103—108.

Renzi, L., 1976b. "Grammatica e storia dell'articolo italiano", SGI 5: 5—42.

Reichenbach, H., 1947. Elements of Symbolic Logic (New York: Macmillan).

Salviati, L., 1586. Del secondo volume degli Avvertimenti della lingua sopra il Decamerone (Firenze:
Giunti).

Soave, F., 1815. Gramatica della lingua italiana. Edizione corretta e migliorata. (Napoli: Nella Stamperia
del Giornale delle Due Sicilie).

Trabalza, C., 1908. Storia della grammatica italiana (Milano: Hoepli).

Yvon, H., 1955-1956. “La notion d’article chez nos grammairiens”, Le Francais Moderne 23: 161--
172, 241--255; 24: 1-13.

Zotti, R., 1836. Grammaire italienne et francaise (Londres: Dulau).

Old High German ä-, Old English &-: A problem that won't go away

Albert L. Lioyd (University of Pennsylvania)

1. The problem

Throughout the entire Old Germanic speech community we find a prefix that may be re-
constructed as *uz-! and occurs in combination with verbs, nouns, and adjectives. Starting
from an apparent original (Germanic) meaning ‘out’ (cf. the identical preposition Gothic
us, Old Norse or, or, ur, yr, Old High German wr, ar, ir ‘out of, from’), it has acquired a
variety of meanings ranging from the spatial (as in ‘go out’), through the actional? (as in
‘carry out’ = ‘accomplish’), to the privative or negative (as in ‘out of luck’ = ‘luckless’).
This prefix appears in Gothic as us-, uz-, ur- (before r) and in Old Norse as or-, ur-, ór-
(here predominantly as a nominal prefix, since weakly stressed [verbal] prefixes were
generally lost; see Heusler 1967: § 123ff.).

In West Germanic, the situation is far more complex. There are obvious cognates of the
North and East Germanic forms: OE or-, OFr., OS ur-, or- (only in nominal composition),
OHG ur- (in nominal composition), ur-, ar-, ir-, er-? (in verbal composition). Beside these,
however, there appear prefixes without - but with similar — sometimes seemingly identi-
cal — meanings: OE, OFries., OS Z- in verbal composition, OE &-, OFries. è, OHG â- in
nominal composition. The distribution in Old English and Old High German, the West Ger-
manic languages in which the great majority of these compounds occur, may be illustrated
as follows:

Noun-compounds or- (or-wene ‘hopeless’) ur- (ur-wäni ‘hopeless’)
æ- (æ-bylg ‘anger’) â- (d-bulgi ‘anger’)
Verb-compounds — 4- (a-belgan ‘toanger’) ar-, ir- (ir-belgan ‘to anger’)

This rather neat scheme is somewhat misleading, however. Indeed, even within its confines
one is struck by the fact that an Old High German verb in ar-, ir- has a corresponding noun
in d-, not ur-; in Old English the reverse is often the case: e.g., or-banc ‘original thought,
genius, understanding’ : a-bencan ‘to think out, devise” (= OHG ir-denken). It is not un-
usual for nominal compounds in or- and @- or ur- and d- to exist side by side, with similar
(identical?) meanings; e.g., OHG ur-herz : d-herz, both glossing ex-cors ‘foolish’; OE or-
gilde : &-gilde ‘without compensation, unpaid for’; or with contrasting meanings, e.g., OHG
ur-werc ‘exercitatio, exercise, practice’ : d-werc ‘stuppa, tow oakum (that which is sepa-
rated out in the course of work)’; ur-kust ‘astus, cunning, cleverness’ : d-kust ‘vitium, fault,
sin, wickedness'; ur-chóse (Notker) ‘indubitable, without contradiction' : d-chòsunga (Gl.)
‘deliramentum , senseless talk’.

244

There are two possible explanations for such an apparently chaotic distribution of forms:
(1) all prefixes are somehow derived from the same Proto-Germanic form, *uz- or some-
thing similar, or (2) we are dealing with two originally distinct and unrelated prefixes, the
semantic fields of which have largely merged. Either solution has certain drawbacks. The
first presents insoluble phonological problems, while the second presupposes the existence
of a Germanic prefix *@- which, it has been claimed, is not only limited to West Germanic,
but has no satisfactory etymology.*

2. The solution — or so it seemed

Jacob Grimm believed in the original identity of all the prefixes listed above and derived
them from earlier *zs- (1870—1878: 1,695f, 781; earlier editions 705, 791). Incidentally,
he incorrectly equated OHG d- and OE a- (which he considered to be long), noting that in
Old High German this form of the prefix occurred only before nouns, in Old English only
before verbs; there is no mention of OE z-. Although Grimm gave five reasons for his con-
clusion, all involved meaning and usage; no attempt was made to clarify the phonological
muddle. Already Graff (1834—1846: I, 16f.) saw objections to identifying d- with ur-, ar-,
etc., but it was Johannes Schmidt (1883: 41f.) who, with his usual incisiveness and neo-
grammatical precision, demolished Grimm’s conjectures and proposed a possible etymology
for the prefix represented in OHG by 4-, in OE by @-. His arguments are worth citing at
length:

“Grimm. . . leitet dies a aus ar- oder as- und identificiert es mit dem as. ags. @-, welches nur
in verbalzusammensetzungen erscheint, beides irrig. Gegen die herleitung von 2 aus ar be-
merkt schon Graff mit recht, dass # nur in nominalzusammensetzungen steht, welche gar
nicht ar, sondern nur ur- gehabt haben könnten..., und hebt begriffsdifferenzen wie
ateilo expers: urteil judicium, achosunga deliramentum: urchosi elucidum hervor. Das as.
ags. d- vor verben ist allerdings aus *az = ahd. ar- entstanden..., aber nicht dem ahd. a
vor nomina gleich zu setzen.”

He then gives examples of the distribution of the prefixes (similar to those given here,
above), and proceeds to remark:

“Die beschrinkung von 4- auf verba, @- auf nomina ist in den poetischen denkmälern streng
inne gehalten, nur ganz vereinzelt überträgt sich das 4- vom verbum auf nebenliegende
abstracta. .. Diese differenz von ags. æ- = ahd. 4- und ags. d- = ahd. ar- sichert als gemein-
sam westgerm. a- (alter €) vor nomina und erweist, dass dies a- nicht aus ar-, der unbetonten
form zu hochtonigem wr-, uz- entstanden ist.”

Schmidt also notes for the first time the existence of an ablaut variant of the prefix *@- in
both OE and OHG: *o-, for example in OHG uo-quemo : d-quemo ‘descendant’; ua-wahst
‘incrementum, stirps, growth, increase, sprout’ : d-wahst recrementum, weed’ (cf. OE
ö-westm ‘shoot, sprout’), etc.

He finds parallels in Latin 4 = ab, abs (which is surely incorrect) and in Sanskrit 4, which
among its many meanings includes that of ‘away from’ (with ablative); an etymology which
was accepted by Walde-Pokorny, Pokorny, and most other standard handbooks. We shall
have more to say about this below.

245

Schmidt’s solution, with later minor corrections and additions, has stood ever since as the
communis opinio, with only two significant challenges, one nearly eighty years ago, and
the other quite recent. Since this latest challenge coincided with the preparation of the
article @-’ for the forthcoming Etymologisches Wörterbuch des Althochdeutschen? (for
which the recipient of this volume serves as a valued consultant), there is a dual reason for
reopening the entire question at this time and reassessing the standard view — which is,
after all, over 100 years old.

3. The first challenge

W. Lehmann (1906) relied on a massive compilation of forms to demonstrate supposed
identity of the prefixes on the basis of usage and meaning, “die Koexistenz des a-, ur-, so-
wie ihre Bindung mit ein und demselben Grundworte” (1906: 34). Only secondarily was he
concerned with justifying this conclusion by attempting to find a phonological explanation
of the various forms. On the basis of two or three exceptional forms in the Old High Ger-
man Glosses in which the nominal prefix appears as ar- instead of ur- and the varying
vowels of the Old High German preposition ur, ar, ir, Lehmann assumes the existence of
nominal compounds with either strongly or weakly stressed prefixes, having developed
from combinations of noun plus strongly or weakly stressed preposition, depending on
sentence stress. He thus derives OHG 4- from the weak form of the prefix (ar-), with subse-
quent loss of r and compensatory lengthening. OE @- is similarly explained, except that
ar > aer before the loss of r (Lehmann 1906: 34ff.).

Apart from his highly questionable theory of the origin of noun compounds, Lehmann in-
volves himself in a number of contradictions: (1) the weakly stressed prefix somehow
became d. in OHG, Œ- in OE before nouns, but ar-, ir- in OHG, a- in OE before verbs (his
explanation: the verbal compounds were formed at a rather late date, at which time the
strongly stressed adverb ur found itself placed in unstressed position! (1906: 79ff.); (2) loss
of r with compensatory lengthening, though common in Old English, is otherwise unknown
in early Old High German; (3) at the time of the change of Anglo-Frisian a > @ (according
to Luick 1921-1940: $ 350, in the 3rd or 4th century), or (< *ur) would not yet have
become ar in positions of weak stress (according to Luick $ 350, this change took place in
the 7th/8th centuries). As might be expected, such a phonological explanation did not
seriously threaten the prevailing view and has fallen into well-deserved oblivion.

Nevertheless, Lehmann’s treatise raises an important issue that should be addressed at this
point: to what extent can semantic criteria (in the broadest sense, including usage) be
relied on to solve this problem — or similar problems?

Those who wish to prove the ultimate identity of the prefixes stress the fact that there are
numerous examples of both 4- (@-) and ur- (or-) attached to the same word, with similar or
identical meanings (see above) and that (rarely) a compound in or- in OE may correspond
to one in 4- in OHG (e.g., or-dale : á-teili ‘not participating in’); those who believe in two
separate prefixes stress the (relatively few) differences in meanings of d- (@-) and ur- (or-)
compounds (see above). In fact, synonymous doublets such as ur-herz : d-herz, or-gilde
: £-gilde could equally well represent alternate forms of the same original prefix (resulting
from varying stress patterns?) or the merging of the semantic fields of two originally dis-
tinct prefixes. This latter interpretation, to be sure, would best account for the differences

246

in d-werc : ur-werc, etc. — differences that Lehmann saw no way to deal with, other than
to deny their importance (1906: 34) and ignore individual cases (Lehmann mentions
specifically only the contrast d-teilo : urteil, which he believes he can discredit as evidence
in the case [p. 42 fn. 3]).

Perhaps a complete ‘field-of-meaning’ study of the prefixes, of the sort pioneered by Jost
Trier (1931) (see also Lloyd 1961), could reveal further differences, though there is little
doubt that most meaning-variants seem to be shared by d-/æ- and ur-/or-. One exception
deserves to be mentioned, however. It has long been recognized that wr-/or- can be a pure
intensifier, without any negative sense, indicating ‘to the furthest degree possible’; e.g.,
OHG ur-alt, OE or-eald ‘very old’ (cf. OHG ir-alten “to age’), OHG ur-märi “famous, out-
standing’, etc. No such compounds with d (Œ-) are attested. It has been claimed that this
function is a later secondary development (Wilmanns 1893-1909: 2, $ 417, 3; Lehmann
1906: 31f.), but that seems highly unlikely, since it rests on one of the original basic
meanings of Germanic *uz- in verbal compounds to indicate a finitive actional type (cf.
Gothic us-fulljan ‘to fulfill, fill completely’) and is found (contrary to Wilmanns) already
in Gothic, where it still shows a connection with an underlying verbal concept; e.g., us-
waurhts ‘righteous, just’ as contrasted with fra-waurhts ‘sinful’ (cf. us-waurkjan ‘to ac-
complish, achieve’). This more positive sense also appears in OHG urwerc ‘exercise,
practice’ (cf. urwercman ‘artifex, craftsman’) as opposed to dwerc ‘tow, oakum’ (see above).

Such differences can certainly be explained most easily if one assumes that d-(@-) goes
back to a particle meaning ‘away from, off’, which never had an actional function as *uz-
demonstrably did. Thus semantic considerations tend to support the standard view, though
they can hardly be said to be conclusive. It seems that it is only in the phonological arena
that any definitive outcome is possible, and until recently the Schmidt hypothesis seemed
unassailable there.

4. The second challenge

In an article (1979a: 30ff.) Alfred Bammesberger took up the gauntlet, challenging the
prevailing view and attempting to demonstrate on the basis of phonological evidence that
Grimm was right after all (also briefly in Bammesberger 1979b). His basic arguments may
be summed up as follows:

(1) PGme. *@-7, assumed by Schmidt and others, has no satisfactory etymology and is
restricted to West Germanic.

(2) PGmc. *@- cannot be the source of both OHG 4- and OE, OF r. a-.

(3) PGmc. *2- cannot be the source of OE @-, because then the prefix would appear as
@- only in West Saxon, é- in the other dialects, which it does not.

(4) Therefore, PGmc. *@- did not exist.

(5) The ‘-losen Formen des Westgermanischen” are derived from *uz-, which became
or-; then proclitic o- became a-; r was lost with compensatory lengthening.

(6) OE z- probably arose by i/j-Umlaut, beginning in the numerous "participia necessitatis"
of the type @-bare ‘manifest, notorious’ (< *uz-berija-) and extended by analogy.

247

(7) Certain parallels can be found in the development of the prefix *fuz-, which appears
both with and without + in OHG zur-, za-, zi-, prep. zuo, za, zi, ze; OE tor-, t0-, prep.
to, te.

(8) The prefix *ö- (OHG uo-, OE ö-) is also derived from *uz-.

The first of these eight points has some validity. Particles, especially those consisting of a
simple phoneme, are notoriously hard to pin down etymologically, and this one is no ex-
ception. There are, however, numerous words in every language without satisfactory
etymologies, and their existence is not challenged. Let us, therefore, postpone the discus-
sion of the etymology of *@- until later, and first dispose of the arguments against its
existence.

First of all, Bammesberger has set up a straw man and then proceeded to knock him down.
By throwing together the OE, OFr., OS verbal prefix 4-8 and the nominal prefix OHG d-,
OE @-, he can safely say (1979a: 32f.) that these cannot all be traced back to “WGmc.
a-”, i.c., PGmc. *@-. It is doubtful if anyone today would make such a phonologically
unjustifiable claim; certainly Schmidt did not. He clearly stated (Schmidt 1883: 41; see
also above) that OS, OE a- before verbs = OHG ar- and has nothing to do with OHG d.
OE z-?

Secondly, Bammesberger is quite correct in stating that "WGmc. *2" occurs in OE as Æ
in West Saxon ande elsewhere. The lack of e- would indeed be unexplainable; however,
there is no lack of e-. Since most extant Old English manuscripts are either West Saxon or
strongly influenced by West Saxon (see Sievers-Brunner 1965: 8 2, especially Anm. 6 and
7), it is no surprise that the overwhelming majority of the occurrences of the prefix have
the form @-, but even a cursory check of Bosworth-Toller (1898—1921) and Wright-
Wiilcker (1884) reveals the following compounds with é- beside the more usual æ- (dialects
are noted where fairly certain): é-bylgdu (Mercian), é-cambe (?), e-mod (?), é-spryngum
(Mercian), é-swic (Mercian), e-uuerdlu (Northumbrian), e-wyrdlu (Mercian). As far as I can
tell, the occurrences of @-are in MSS with at least some West Saxon traits.

There is thus no reason to deny a PGmc. *@-as the source of OHG 4-, OE @- (é-) — and also
OFr. e-, which was not mentioned by Bammesberger, but occurs in the compounds e-bete
‘unatoned’ (‘busslos’), é-felle ‘without skin or covering’ (cf. OE @-felle), é-fréthe ‘without
paying Friedensgeld’ (a fine paid to the authorities; see Hoops 1911—1919: 2, 96ff.),
é-live ‘lifeless’.'° In fact, as we shall see, no other source can plausibly be posited.

Bammesberger's fifth point is one I would agree with completely, if by the “r-losen For-
men des Westgermanischen” we understand the verbal prefix #- in Old English, Old Frisian,
Old Saxon." For Old English, each step of the development represents a standard pho-
nological change:

(1) *uz > *ur (standard West Germanic rhotacism; Sievers-Brunner § 181, 2).
(2) ur > or (u > o before r <z; Sievers-Brunner § 45, 5).

(3) r was lost in monosyllabic words (Sievers-Brunner $ 182). This loss, which was stan-
dard in Old English, Old Frisian and Old Saxon — the languages that have a verbal prefix
without r — did not occur in (early) Old High German — which has a verbal prefix with
r —; compare for example OE me : OHG mir ‘to me’; OE hva : OHG hwer ‘who’. This is
another reason to rule out the derivation of OHG d@- from *uz-.

248

After the loss of r, the remaining vowel was lengthened in words that could be used in
strongly stressed positions, but probably remained short when such words occurred with
weak stress (Sievers-Brunner $ 137,1 and Anm. 1). The length of the vowel in the verbal
prefix a- is argued. Since the loss of r must have occurred while *or was still an independent
preposition, able to appear in various stress patterns (cf. OHG ur, ar, ir), it is possible that
both long and short vowels occurred, but since verbal prefixes were regularly weakly
stressed, I would agree with Lehmann (1906: 80, n. 2) in positing a short vowel for such
prefixes.

(4) *o- 7 a- in positions of weak stress (Luick 1921—1940: 8 323).

While this explanation deals nicely with the verbal prefix a-, it cannot be applied to OHG
d-, as we have seen, nor does it yield an OE &-. Bammesberger’s solution to the problem of
this troublesome form is ingenious and even theoretically possible, though it does require
considerable analogical extension to forms that contained no umlaut-producing sounds.
One might ask, however, why such an umlaut occurred only in nominal compounds; a
number of compounds in a- with first-class weak verbs are equally capable of umlaut (e.g.,
a-firran, a-lvsan, a-lecgan, a-pencan, etc.); or why the numerous nominal or- compounds of
the type or-dwle did not generate umlaut of this prefix. One could extend such analogies
at will. Obviously the fact that umlaut did not take place in other similar compounds does
not rule out the possibility that it did, for whatever reason, in @- compounds. This is the
sort of speculation that cannot be disproved; however, it is fortunately unnecessary in
order to explain the Old English prefix &-.

Bammesberger's final two points are in fact rather peripheral to the central argument, and a
detailed discussion of the myriad problems connected with the prefix *tuz- and other
related and non-related forms could provide the matter for another entire article. Suffice
it to say that at least two and possibly three different word groups have been thrown
together here. To single out the one nearly certain thing in the mass of uncertainties
surrounding these words, OHG zuo, OE to ‘to’, both also used as prefixes, are totally un-
related to OHG zur-, OE tor- (nominal prefix) ‘bad, difficult, asunder’, OHG zar-, zir-, za-,
zi-, OE to-, te- (verbal prefix) ‘apart, asunder’. Whether in turn OHG zur-, OE tor- are
really identical to OHG zafr)-, zifr)- (the r may be secondary, from ir-), OE #0-, te-, or
whether the former are connected with Gothic tuz-, Greek 5vo- and the latter with Lat.
dis-, is impossible to say for sure.'? In any case, this group of words provides rather shaky
ground on which to erect any edifice. Moreover, I fail to see any real parallel to ur- : á-.
Even if zur-/tor- is the stressed equivalent of za/r)-/t0-, it consistently retains the r, just like
ur-/or-, and never loses it to yield a form corresponding to d-/@-. The usual loss of r in the
Old High German verbal prefix za-, zi- does not agree with its consistent retention in OHG
ar-, ir-.?

Finally, not only is the attempt to derive the prefix uo-/ö- from *uz- phonologically
questionable, especially for Old High German, where no parallel exists for early loss of r
with compensatory lengthening, but Bammesberger also fails to explain why the same
nominal prefix *uz-, which already is supposed to appear sometimes as ur-/or-, at other
times as d-/æ-,sometimes takes on still a third form uo-/o- — even alternating with 4-/z- in
the same word, as in OHG d-quemo : uo-quemo. Here varying stress patterns simply won't
work.1^

249

Since not a single argument against the derivation of the prefix OHG d-, OE æ- (é-), OFr.
è- from PGmc. *- can be upheld, and the arguments for its derivation from PGmc. *uz-
are phonologically implausible, there would seem to be no reason to question the existence
of a PGmc. particle *2-,as accepted in most handbooks.

5. The etymology of PGmc. "z-

To return to the first point raised by Bammesberger, there is no doubt that etymological
support for Gmc. *@-is weak. It is true that the particle can be identified only in West Ger-
manic, where, it is interesting to note, it seems to be in decline, eventually dying out in all
languages, with the exception of a few isolated remnants like NHG (dialectal) Öhmd
‘aftermath, second mowing’ (< d-mád). It is totally lacking in Gothic. In Old Norse, the
situation is not so simple. If the prefix had survived, it would have fallen together with the
prefix d- = Gothic ana-, German an-, and indeed in some environments with d- < af- (see
Noreen 1923: 8 299,5; Cleasby-Vigfusson 1957: 5 [s.v. af, DJ), which could well have been
sufficient to hasten its final disappearance. G. Schmidt (1962: $ 176) believes remnants of
an original á- < *æ- can still be identified in a few compounds; he lists a-mättigr, an epithet
of witches and giants, which he interprets as ‘with evil power” (Cleasby-Vigfusson [ 1957:
43] even compares OHG d-mahtig, which, however, means ‘powerless”), d-verk{i) ‘unlawful
work’, d-voxtr ‘fruit, produce, growth’ (OHG d-wahst means ‘weed’, but compare the
ablauting uo-wahst ‘growth, increase, sprout’), d-vdn, d-vaeni ‘faint expectation’ (cf. OE
@-wen ‘doubtful, uncertain’), d-munr ‘eager (for revenge)’ (contrast d-munr ‘like, similar’
< *ana-).

None of these must be traced back to an original *z- — indeed, I have my doubts about a
number of them —, but they indicate that we simply have no way of knowing whether Old
Norse once had such a prefix or not, and that its supposed isolation in West Germanic may
not have been so complete.

Before leaving the Germanic languages, a word must be said about the prefix that appears
in OHG as uo- and in OE as ö-. As we have seen, it can have the same meaning as *@-
(äquemo : uoquemo) or a different meaning (dwahst : uowahst). Attempts by Lehmann
(1906) and Steinhauser (1960) to demonstrate that *o- must be unconnected with *2- on
the basis of semantic differences are not fully convincing. To be sure, the few examples of
*ó-compounds that we possess show a smaller range of meanings than those with *@-
— there are no examples of *o- with a purely negative meaning, for example, unless OS
Óbult belongs here (see below) —, but *o- expresses no meaning not shared, or at least
approximated, by &-. Typical! meanings of *o- are: ‘down (away from)’ as in OHG uo-hald
‘precipitous, sloping down’; ‘after, later” as in uo-quemo; and (closely related to this
meaning) ‘added, supplementary’ as in OE 0-westm, perhaps also in 0-web ‘subte(g)men,
woof (in weaving)’. All of these meanings can easily be derived from ‘proceeding away
from’, and the last two can be compared with the use of *@-in d-quemo, á-mád. While such
evidence is not sufficient to prove anything conclusively, the existence of two such similar
particles certainly suggests the possibility of ablaut variants, as J. Schmidt claimed. Unless
a better explanation can be found, this can at least be accepted as a working hypothesis.

The particle *o- is quite rare, with certain occurrences only in Old English and Old High
German.! It may be found in Old Saxon in two words: Gbult ‘anger’ (Wadstein 1899:

250

52.212) and ölat ‘thanks’ (Heliand, with variant alat). The first may equally well be an
error for *orbul/h}t (cf. Holthausen 1921: 8 227 Anm.), while the second has no sure
etymology, but has been connected with Gothic awi-liup ‘thanks’ (Feist 1939: 70; Berr
1971: 308).*© The often cited MLG word 6-herde (incorrectly listed by Steinhauser [1960:
101] as Anglo-Saxon), usually translated ‘Unterhirt’, is probably actually ‘sheep-herdsman’,
from ouwe, owe ‘sheep’, as in the compound ö-höf ‘sheep farm’ (Schiller-Lbiben 1875—
1881: 6, 226f.). There is no sign of the prefix in Old Frisian, or in Gothic. In Old Norse,
there is again no way to determine whether the prefix was present, since it would have
fallen together with the much more common negative prefix ó- « *un- *un-' (see Noreen
1923: 8 112,1). Either this would have resulted in contradictory pairs (*6-hallr ‘preci-
pitous’ < *Ö- : ö-hallr ‘not sloping’ < *un-), or, if *o-, like *@- ever developed a negative
meaning, this function could have been taken over by 6-< *un-!” (ON ó-baettr *unatoned'
is synonymous with OFr. e-bete).

The most obvious cognate to Gmc. *@- and *o- in non-Germanic languages is the ubi-
quitous Indo-Iranian preposition and postposition (with accusative, locative, ablative) and
prefix a(-), which has a large array of meanings: ‘to, up to; in, among; in addition to;
somewhat; away from'.!? It has been suggested that all these meanings arose from the basic
idea of ‘near’ (Pokorny 1959: 280; Steinhauser 1960: 103), which, in order to arrive at
‘away from’, would apparently require a development via ‘in the vicinity of, but no longer
at’. This is certainly possible; the history of particles is frequently one of wide-ranging
semantic change. As particles develop less specific functions, however, there is always the
possibility of merger of similar particles, and this could also be the case here. One could
theoretically posit a merger of particle, with a basic meaning ‘to, at, in’ or the like, and
particle, ‘away from’, which could easily overlap in borderline areas such as ‘near’, ‘some-
what’. I leave it to specialists in the Indo-Iranian languages to choose which alternative best
fits the history of the particle in those languages.

In any case, Indo-Iranian a can represent PIR zo, *e, or *0. The Germanic particle *@- can
only come from PIE *e; the particle *0- could come from either *o or *a, though an ablaut
pair *é- : *o- is more probable. Moving beyond Indo-Iranian, we find in the Slavic lan-
guages a rather rare ‘colorless’ prefix ja-, as in Serb.-Church Slav. ja-skud» beside skod»
“ugly(?)’,'” which may be related — and which again could come from PIE *4, Se, or SO
Evidence for *0 may be provided by one or two Greek words: xnpworat ‘collateral
relatives who inherit for the lack of closer relatives’ (*xNpov verwaistes Gut’ + *6-d-ta; cf.
Skt. a-da- ‘receive’, ppl. a-t-ta- ‘received’; so also Frisk 1960-1972: 2, 1096f.; Mayrhofer
1953—1980: 1, 69; opposed Steinhauser 1960: 104) and possibly &-xp0s ‘yellowish, pale?
(cf. the use of the Sanskrit prefix d- in such words as @-pifigala- ‘yellowish’), though this
etymology is doubted by both Frisk (1960-1972: 2, 1153) and Mayrhofer (1953-1980:
3, 274). Since a connection between Gr. coxpds and Skt. vyaghrá- ‘tiger’ is highly unlikely,
the second element of Gr. wxp0s is left totally isolated.

Evidence for *e is very weak outside Germanic: perhaps ablauting with the © in Gr.
xnpworatis the é in Lat. heres, -edis ‘heir’ (< *gher-e-d-?; accepted by Walde-Hofmann
1938-1956: 1, 641f.; Frisk 1960—1972: 2, 1096f.; opposed Steinhauser 1960: 111) even
more uncertain in Gr. nBaws beside Bats ‘little’ (opposed by Frisk 1960—1972: 1,619;
uncertain Schwyzer 1953: 2,491). There is no specific evidence for "a. Attempts to
connect various particles with the short vowels e and o as ablaut variants are all uncertain
to varying degrees (see Pokorny 1959: 280f.).

251

The overall etymological evidence is not as strong as one might wish, but at least there
seems no reason to question the connection of the Germanic particles with Indo-Iranian
a(-), which is both semantically and phonologically plausible, whether one chooses to
derive the Indo-Iranian particle from PIE *e or *o, or even from a coalescence of two
ablaut variants or two originally distinct particles that came to be felt as ablaut variants.
While all such details may never be entirely clarified, the central fact of the existence of

the Germanic particles "a- and "o- remains as firmly established today as at the time of
Johannes Schmidt.

Notes

Probably from PIE *ud-s, a derivative of the particle *ud- (Skt. úd-, út- ‘up[ward], out; with long
vowel Gothic, OE üt, OHG wz), with an s-determinative (as also probably in the synonymous Skt.
us-, Avestan uz-). See Walde-Pokorny (1927-1932: 1, 189f.); Pokorny (1959: 1103f.); Feist (1939:
528); Kluge (1975: 808) (inconclusive).

For a discussion of ‘actional types’ — similar to what are called Aktionsarten in German, see Lloyd
(1979: 91ff., 103f. [finitive function of Gothic us-]).

ur- is early and rare, er- is late Old High German and Middle High German. The 9th-century East
Franconian form ir- is generally used in examples, unless a specific form is cited.

For a complete list of OHG @-compounds see Karg-Gasterstädt (1952-: 1, 1). OE @-compounds have
been collected from various sources, including Bosworth-Toller (1898-1921) and Lehmann (1906).
Edited by Albert Lloyd and Otto Springer and supported by grants from the National Endowment
for the Humanities and the William Penn Foundation.

Of course the further analogical extension of this prefix to such words as OHG ur-@no ‘ancestor’,
NHG Ur-wald ‘primeval forest’, etc., building on such compounds as wr-alt, is secondary, as is the
development of the meaning ‘original, proto-’.

Bammesberger refers throughout to “West Germanic *4”, a formulation that is of course phonem-
ically every bit as proper, but which, for phonetic reasons, I prefer to avoid. The development of PIE
*ë to OHG, OS 4, OE ae, OFr. é presupposes a PGmc. rather low front vowel in the general area of
[z:] but with higher and lower allophones.

The length of the vowel is disputed; see below.

Bammesberger (1979a: 31, n. 5) misstates Schmidt’s position.

See Helten (1907: 88, 89, 94). The prefix d- in the noun compounds dbél ‘a raised scar’, dbére
beside the more common aubére) ‘manifest’, and ddel ‘legate’ (? listed only in Holthausen 1925: 1)
has been explained variously as an error or an analogical transfer of the verbal prefix. See Helten
(1907: 24, 28).

Lehmann (1906: 79f.) also came to a somewhat similar conclusion, as did Siebs (1899: $ 101, Anm.
2) for Frisian. It is also verified by the rare survival of the full prefix ar- before a vowel in OE ar-efnan
‘perform’, “falsely divided to alliterate on r, and giving rise to a simplex ra@fnan” (Campbell 1959:
§ 73 fn. 2).

On the etymology of OHG zuo, OE t6 < PIE *dé : *dò, see Kluge (1975: 889), Pokorny (1959:
181ff.). On the problematic relationship of the prefixes, see Braune-Eggers (1975: § 72, Anm. 2),
Lehmann (1906: 80ff., n. 2); Kluge (1975: 880) (merges OHG zur- with za[r]-, zi[r]-, with no
mention of OE tor-). Most handbooks separate the two sets; e.g., Feist (1939: 119 [s.v. dis-), 484
[s.v. tuz-werjan]); Holthausen (1963: 350 [td-, te-], 351 [tor-]); Pokorny (1959; 232); Walde-Hof-
mann (1938-1956: 354£.); Frisk (1960-1972: 1, 425).

The much less common Old High German forms with - were limited to Upper German; only in
Middle High German did r-forms become more frequent, in New High German entirely supplanting
the others (see Braune-Eggers 1975: 8$ 72 and Anm. 1, 2). The different behavior of ar- and za(r)- has
not been satisfactorily explained. G. Schmidt (1962: 8 312) plausibly suggests that it may result
from the existence of the former as an independent preposition (regular retention of -r in mono-
syllables with short vowels), as contrasted with the sole use of the latter as a weakly stressed prefix
(loss of -r — or perhaps already *-z — in pretonic position?).

252

M I do agree with Bammesberger, however, in connecting uo-/ö- with d-Jæ-, contrary to the views of
Lehmann (1906: 138ff.) and Steinhauser (1960).

Since the occurrences are so limited, a complete list might be useful (Old High German citations are
from my files, Starck-Wells (1972—), and Schützeichel (1981); spellings are normalized; Old English
citations rely on Bosworth-Toller, so completeness is not guaranteed):

OHG: uo-hald/i), -ig ‘precipitous, sloping down’; uo-haldi ‘precipice, steep slope": uo-kalo 'recalvaster,
bald in front, with a high bald forehead’ (perhaps actually ‘balding’; cf. Skt. d- ‘somewhat’, and note
that Lat. recalvus, closely related to recalvaster, can mean ‘bald in front’ or ‘bald in back"!); uo-
kumft ‘succession’; uo-kumil{o), -quemil{o), -kumiling ‘late-ripening grape’; uo-quemo ‘descendant’;
uo-staft ‘pittacium, patch’; uo-wahst ‘growth, increase, sprout’; uo-zurnen, -zarnen ‘to disdain, scorn’
(from a noun *uo-zorn?). The often cited OHG *uo-mdd does not exist; see Steinhauser 1960: 101.
MHG: uo-sedel ‘vectis, lever or bar’; uo-sezzel (?) ‘something to cover a hole in a cloak’ (Lanzelet
6023; see Lexer 1872—1878: 2, 1998; Benecke-Miiller-Zarncke 1854-1861: 2, 2, 339; the MS has
vosezzel, which could also stand for *vor-sezzel. See also Pfeiffer Germania 3 (1858): 480, who
suggests "für-fezzel; unlikely). OE: o-den ‘threshing floor’; 6-gengel ‘bolt (for a door)’; ö-heald, -hilde
‘precipitous, sloping down’; 6-leccan ‘to flatter, be obsequious’, ö-leccere ‘flatterer’, ö-leccung
‘flattery, allurement’ (all probably from an original lost noun-compound; cf. also @-lecung ‘allure-
ment’; I can find no trace of an OS ö-leccan, cited by Steinhauser 1960: 102); 6-weestm ‘sprout,
shoot’; 6-web, -wef 'subte(g)men, woof'.

Steinhauser (1960: 102) also connects Old High German proper names in Ô., such as Ó-gast, Ua-gast,
Ô- hilt(a), Ó-munt, Ó-reht: the etymology of such names is, however, extremely uncertain.
Steinhauser's (1960: 102) connection with MHG letzen is improbable, since the basic meaning of
this word (as also OHG lezzen) was ‘hinder’ (cf. Engl. let). A direct connection with OS /dtan ‘let,
allow, leave’ suggests itself immediately, but presents severe semantic difficulties (‘a farewell gift'?).
Compare such Old High German synonyms as d-kust and un-kust, d-mahtig and un-mahtig.

18 See Bôhtlingk-Roth (1855-1875: 1, 581ff.); Monier-Williams (1899: 126). Unless otherwise spe-
cified, the following etymological data are from Walde-Pokorny (1927-1932: 1, 95£.) and Pokorny
(1959: 280f.). See also the forthcoming Etymologisches Wörterbuch des Althochdeutschen, sv. d-.
Rozwadowski (1909: 102); cf. Old Church Slav. skodb ‘needy, wanting’, but Polish pa-skudny
‘foul, nasty, ugly’.

References

Bammesberger, A., 1979a. “Die westgermanischen Entsprechungen zu urgerm. *uz(-)”, BGDSL (Tübin-
gen) 101: 30-35.

Bammesberger, A., 1979b. Beiträge zu einem etymologischen Wörterbuch des Altenglischen (Heidel-
berg: Winter).

Benecke, G.F./W. Müller/F. Zarncke, 1854-1861. Mittelhochdeutsches Wörterbuch (Leipzig: S. Hirzel).

Berr, S., 1971. An Etymological Glossary to the Old Saxon Heliand (Bern/Frankfurt: Lang).

Böhtlingk, O./R. Roth, 1855-1857. Sanskrit-Wörterbuch (St. Petersburg: Kais. Akad. der Wissenschaf-
ten).

Bosworth, J./T.N. Toller, 1898--1921/1972. An Anglo-Saxon Dictionary (With enlarged Addenda and
Corrigenda to the Supplement by A. Campbell, 1972) (London: Oxford University Press).

Braune, W./H. Eggers, 1975. Althochdeutsche Grammatik (13. Auflage) (Tübingen: Niemeyer).

Campbell, A., 1959. Old English Grammar (Oxford: Clarendon Press).

Cleasby, R./G. Vigfusson, 1957. Icelandic-English Dictionary (2nd edition) (Oxford: Clarendon Press).

Feist, S., 1939. Vergleichendes Wörterbuch der gotischen Sprache (3. Auflage) (Leiden: Brill).

Frisk, H., 1960—1972. Griechisches etymologisches Wórterbuch (Heidelberg: Winter).

Graff, E.G., 1834-1846. A/thochdeutscher Sprachschatz (Berlin: Nikolai'sche Buchhandlung).

Grimm, J., 1870-1898. Deutsche Grammatik (Neuer vermehrter Abdruck besorgt durch W. Scherer,
G. Roethe und E. Schröder) (3. Auflage) (Berlin/Gütersloh: F. Dümmler).

Helten, W.L. van, 1907. Zur Lexicologie des Altostfriesischen (VKNA N.R. 9) (Amsterdam: J. Müller).

Heusler, A., 1967. Altisländisches Elementarbuch (7. Auflage) (Heidelberg: Winter).

Holthausen, F., 1921. Altsáchsisches Elementarbuch (2. Auflage) (Heidelberg: Winter).

253

Holthausen, F., 1925. Altfriesisches Wörterbuch (Heidelberg: Winter).

Holthausen, F., 1963. Altenglisches etymologisches Wörterbuch (2. Auflage) (Heidelberg: Winter).

Hoops, J., 1911—1919. Reallexikon der germanischen Altertumskunde (Straßburg: Trübner).

Karg-Gasterstädt, E./Th. Frings et al., 1952. Althochdeutsches Wörterbuch (Berlin: Akademie-Verlag).

Kluge, F., 1975. Etymologisches Wörterbuch der deutschen Sprache (21. Auflage) (Bearbeitet von
Walther Mitzka) (Berlin: de Gruyter).

Lehmann, W., 1906. Das Präfix uz- besonders im Altenglischen (Kieler Studien zur englischen Philologie
N.F. 3) (Kiel: R. Cordes).

Lexer, M., 1872—1878. Mittelhochdeutsches Handwórterbuch (Leipzig: S. Hirzel).

Lloyd, A.L., 1961. “The Verbs of the Meaning-Class ‘Do or Make’ in the Old High German of Notker
Labeo”, GR 36: 245-256.

Lloyd, A.L., 1979. Anatomy of the Verb: The Gothic Verb as a Model for a Unified Theory of Aspect,
Actional Types, and Verbal Velocity (Amsterdam: Benjamins).

Luick, K., 1921-1940. Historische Grammatik der englischen Sprache (Leipzig: C.H. Tauchenitz).

Mayrhofer, M., 1953-1980. Kurzgefaßtes etymologisches Wörterbuch des Altindischen (Heidelberg:
Winter).

Monier-Williams, Sir M., 1899. A Sanskrit-English Dictionary (2nd edition) (Oxford: Clarendon Press).

Noreen, A., 1923. Altisländische und altnorwegische Grammatik (4. Auflage) (Halle: Niemeyer).

Pokorny, J., 1959—1969. Indogermanisches etymologisches Wörterbuch (Bern/München: Francke).

Rozwadowski, J., 1909. “Review of Berneker, Slawisches etymologisches Wörterbuch”, Rocznik
Slawistyczny 2: 57—112.

Schiller, K./A. Lübben, 1875-1881. Mittelniederdeutsches Wörterbuch (Bremen: J. Kühtmann).

Schmidt, G., 1962. Studien zum germanischen Adverb (Dissertation) (Berlin).

Schmidt, J., 1883. ‘Die germanischen Präpositionen und das Auslautsgesetz”, Zfvgl.Spr. 26: 20-45.

Schützeichel, R., 1981. Althochdeutsches Wörterbuch (3. Auflage) (Tübingen: Niemeyer).

Siebs, Th., 1899. “Geschichte der friesischen Sprache”, in: Grundriß der germanischen Philologie 1
(2. Auflage), 1152-1464.

Sievers, E./K. Brunner, 1965. Altenglische Grammatik (3. Auflage) (Tübingen: Niemeyer).

Starck, T./J.C. Wells, 1972. Althochdeutsches Glossenwörterbuch (Heidelberg: Winter).

Steinhauser, W., 1960. “Die althochdeutsche Vorsilbe uo- im Kreise ihrer Lautverwandten”, Zeitschrift
für Mundartforschung 27: 101—115.

Trier, J., 1931. Der deutsche Wortschatz im Sinnbezirk des Verstandes (Heidelberg: Winter).

Wadstein, E., 1899. Kleinere altsächsische Sprachdenkmäler mit Anmerkungen und Glossar (Leipzig/
Norden: D. Soltau).

Walde, A./J. B. Hofmann, 1938-1956. Lateinisches etymologisches Wörterbuch (3. Auflage) (Heidel-
berg: Winter).

Walde, A./J. Pokorny, 1927—1932. Vergleichendes Wörterbuch der indogermanischen Sprachen (Berlin/
Leipzig: de Gruy ter).

Wilmanns, W., 1893—1909. Deutsche Grammatik (Strafburg: Trübner).

Wright, T. / R.P. Wülcker, 1884. Anglo-Saxon and Old English Vocabularies (London: Trübner).

Quelques noms grecs à l’Agora d’Athènes

Olivier Masson (Université de Paris X, Nanterre, et Ecole des Hautes Etudes, Paris)

En 1976, Mme Mabel Lang a donné une excellente publication des graffites et dipinti qui
ont été rassemblés à la suite des fouilles américaines de l’Agora!. La consultation de ce
travail fournit l’occasion de revenir sur un certain nombre d’anthroponymes plus ou moins
rares.

Je signalerai tout d’abord à l’attention le seul document qui, quoique pleinement grec,
n’est pas rédigé en alphabet, à savoir un petit graffite incisé sur le pied d’une kylix à vernis
noir (deuxième partie du V® siècle). Fait notable, mais pas trop surprenant, il s’agit de cinq
signes syllabiques chypriotes, no. F 67 (Plate 13, dessin). L'éditeur a correctement dessiné
et transcrit: ku-po-ro-fa-mo où Kvrpoôduw. Gravure en syllabaire commun, de droite à
gauche, avec des signes assez réguliers (selon le dessin); noter seulement, à gauche, un mo
de forme presque carrée?. On doit alors ajouter que le composé Kurpó-8agoc, parfaitement
formé, est nouveau, et vient s'ajouter à la riche série des composés constitués avec un
premier élément Kurpo-, série qui est naturellement caractéristique pour Chypre*.

D'ailleurs, la présence de Chypriotes en Attique est déjà attestée par quelques témoignages,
grâce à des noms de dédicants ou de défunts, ordinairement transmis en alphabet. On citera
ainsi les Salaminiens de Chypre ‘EAouérns (fils) et ‘EAAayopas (père), dans une épitaphe
du Pirée, IG 112, 10217/8 (aprés le milieu du IV s.)^. Ou encore, un homme originaire de
Marion, souvent cité? , IIvvró&auoc Xraoayópov, en IG II, 9284. Etc.

Mais ce qui est trés remarquable pour le graffite de |” Agora, c'est l'utilisation du syllabaire
chypriote en Gréce propre: on a ici le second exemple connu, aprés la découverte assez
récente d'un pied de lion en bronze de Delphes, incisé avec quelques signes, avec la dif-
férence que cet objet, appartenant à un trépied, a sürement été importé de Chypre.

J'examinerai ensuite quelques noms qui me paraissent dignes de commentaire, en suivant
simplement l'ordre alphabétique.

BAóovc, sur un fragment de pithos du VIe s., D 29, Pl. 8. Ce nom ne serait pas connu, du
moins en Attique; effectivement, il manque dans les répertoires usuels. En fait, il est attesté
à Abdére, et pourrait bien représenter un nom ionien, quelle que soit l'origine du person-
nage à l'Agora. Une émission rare de monnaies d' Abdére du V® siècle”? porte la légende très
claire &ri BAóovos: on sait que, sur de nombreuses séries d'Abdére et de Maronée, comme à
Byzance, le nom du magistrat monétaire est très souvent indiqué grâce à la formule éri plus
le géngt". La forme BAdouc est évidente, mais comment l’expliquer? Elle ne paraît pas
avoir attiré l'attention de Bechtel. Il me semble, en tout cas, qu'on pourrait l'interpréter
comme un diminutif en -uc? de l'adjectif homérique ffAooupóc, lequel signifie probablement

256

“au regard terrible”, selon l'analyse de Leumann?. L'existence d'un terme de ce genre

comme sobriquet m'aurait certes rien de surprenant. Un autre nom rare BAdoww est pro-
:104

bablement apparenté 1%.

‘Endvaoos, sur lanse d'un petit canthare à vernis noir, daté vers 520—490, G 4, PI. 30.
Ce nom figure dans une dédicace archaïque à Hermès, 'Erdvacos heppei!!. Le premier
nom est curieux, mais parait bien lu, et il ne semble pas qu'on puisse comprendre autre-
ment le début avec un nom simple plus connu, soit EII(?) 'Ovaooc. L'éditrice, sans com-
mentaire, a vu ici un composé nouveau 'Er-óvaooc. Il s'agit alors du groupe des composés
en 'Ovaci-, qui ne fournit pas, en règle générale, de second élément de composé en -óvaooc/
-óvnooc!? . Cependant, il se trouve que j'ai commenté naguére l'apparition d'un composé
fort rare de ce type, à savoir Mopr-óvaooc dans une inscription de Théra, IG XII Suppl.
no. 67 (épitaphe, VE siècle). Ce nom doit représenter le “renversement” des éléments
constituant le composé banal ‘Ovaot-uBporos; dans le cas d’ "Er-övaoos, il faudrait plutôt
songer à l'influence, dans l'onomastique d'une famille, de l'un des trés nombreux composés
en ‘Er(t)-*. On notera enfin que le nom, avec un a long conservé, ne peut guère être atti-
que.

Eürpaxois, sur un fragment de grande amphore du début du VI® siècle, D 7, PI. 7. Lettres
archaïques, graphie xo. L'intérét de ce graffite est de fournir un exemple ancien du nom de
femme Eënpaë, que Bechtel ne pouvait pas citer avant le IV? siéclelé . Ce nom, trés clair,
est simplement le substantif abstrait edmpaëic “bonheur” (variante, employée par Eschyle,
du banal ebnpaëia), et se rattache à une catégorie trés riche de noms féminins; à l'époque
impériale, les noms Eupraxia et Eupraxis sont encore en usage à Rome!?.

M&wos, sur l’anse d’une jarre tardive, V® ou VIF siècle de notre ère, F 325, Pl. 28. Graffite
Ma%kov, assez légèrement gravé. L'éditeur évoque un nom MaAmxos, attesté en Attique
(notamment PA 9661, Acharnien, dans IG I2, 5816). J'ai étudié en détail ce nom ancien
et rare, qui apparait à Corinthe dés le VIII ou VII s. avant notre ére, et a été employé
dans diverses régions, toujours avant notre ére!. Je doute qu'on puisse le retrouver si tard
à l'Agora. A l'époque de ce fragment, je verrais plutót ici une variante du nom sémitique
assez banal MáXxoc, attesté en Syrie, à Palmyre, en Egypte, etc!°; une variante Máukoc
précisément connue à Palmyre?? . D'autre part, le nom est attesté en Attique, au moins sous
l'Empire, à en juger par une épitaphe du Laurion, IG II2, 12029, qui porte MáXxe xaipe?! .

zav9ns. Sur un tesson venant d'une amphore non vernissée, datable vers 490—450, D 39,
Pl. 9, on trouve toute une liste de noms au nominatif, cinq masculins, deux féminins, soit:
Mevéônuos, Xapias, =avôns, AuyBov\oc, Ilowrapxos, Hevrapiorn, KaA(AuoTpärn. La
plupart de ces noms n’appellent pas de commentaire spécial. Cependant le troisième,
Sonne, est notable comme un bon représentant de la catégorie connue des noms ioniens
où -ns résulte de la contraction de -éas. En face de Zavdeas, on a donc Zave, déjà at-
testé mais plus tard, en domaine ionien” .

Hevrapiorn, même liste, sixième nom. Ce féminin est à première vue bizarre, car l'on
attend Ilavrapiorn; cependant, l’epsilon est certain, et un masculin similaire Tlevrapıoros
est attesté plus tard à Cyrène??. D'autre part, une stèle attique (Paiania) du début du IV® s.
offre la double légende Ilevrapiorn Ilavrapiorn selon l’éditeur*. D. Robinson voyait ici
deux noms de femme différents, pour deux. personnages. Le second serait le nom normal
Ilavraptorn. Mais le premier serait autre chose, avec cette curieuse explication:

257

“Ievrapiotn is possibly the name of the child whose mother may have died in child-birth.
She may have been the fifth child and so called the fifth best (sic), or the fifth of the Ariste
family. The name is new, but it is a perfectly possible formation. . DA

Une telle analyse est impossible, et l'on sait qu'il n'existe pas de noms constitués avec un
premier élément Ilevr(e). "cinq". D'autre part, le voisinage des deux orthographes sur
l'épitaphe est remarquable. Je ne saurais dire s'il s'agit ici réellement de deux femmes
différentes, appartenant à la méme famille, ou plutót de deux variantes, le méme nom étant
écrit sous deux formes. En effet, j'estime que IIevrapiorn, à còté du Ilevrapıoros cyreneen,
représente une variante rare et sporadique, dans laquelle le premier epsilon résulte simple-
ment d'une dissimilation phonétique des deux a consécutifs. Un paralléle en Attique serait
fourni par la graphie Veud@n à côté de Yaudôn*. Ainsi, tout rentre dans l'ordre, et l'ex-
emple de 1’ Agora, au début du V° s., est intéressant par sa date.

IIpaxo(t)vn, sur un fragment de grande amphore, VII*/VI® s., D 10, PI. 7. L’interprétation
comme IIpa£ívn, soit le féminin de IIpa&ivoc, est évidente. On remarquera la date haute et
l'absence d'iota qui doit résulter d'une graphie rapide, bien admissible sur un graffito??.

IIUxocow(?). Une lekane du V* s., C 21, Pl. 5, porte des noms et l'acclamation KaAóc, in-
scrits négligement dans un quadrillage. A la 1. 3, l'éditeur suggére de lire IUxocovoc, vrai-
semblablement comme génitif. Si cette interprétation est correcte, je me demande si l'on
n'obtient pas ainsi un sobriquet nouveau, Ilvxowv = IlvEwv, tiré de mvéos “buis”. Comme
on sait, Bechtel a relevé plusieurs dérivés en -wv de ce type, comme Adpvwv, Apvwr,
"EAarwv, Mipowv, etc”®.

DucéAa. Sur la base d’un skyphos du troisiéme quart du V® s., C 27, Pl. 6, on a le graffite
XuéAa kararÿy(awa), qui relève d’un:schéma bien attesté??. Il faut noter seulement le
nom Zwé\a, tiré de l'ethnique correspondant, car il était inconnu de Bechtel, pour le

masculin comme pour le féminin®

[Zov(u)Bons. Fragment inscrit, base d’une coupe à vernis noir, second quart du V® s.,
F 84, PI. 14. Sept lettres régulièrement incisées. Avec une restitution plausible à gauche,
l'éditeur écrit ZJLou(u)Bpes et remarque:

“The eta (written as epsilon) following rho suggests a foreigner. The nearest parallel for this
name is ZxovuBpuv (nick})name for an hetaira in Theophilos’ Flute-lover (Edmonds...,
fr. 11)”

Effectivement, avec une lettre de plus et la non notation de la nasale?! , il est tentant de
retrouver un tel nom. Cependant, on écrira plutót ZovuBpNs, avec une terminaison ionien-
ne du méme type que Zav@Ns étudié ci-dessus. D'autre part, on peut citer aujourd'hui des
exemples supplémentaires de surnoms de cette famille, formés sur otoUufipiov “menthe
aquatique" ou “cresson”**. Bechtel lui-même aurait dû signaler les noms masculins Provu-
Bpas et Zuwuufptokoc, employés pour le méme personnage chez Hérondas, Mime II, 76.
Enfin, il ne faut pas oublier que ces surnoms ne sont pas nécessairement péjoratifs: à
Athènes, on relévera le masculin probable ZiovuBpis, 1G 112, 12615 (majuscules d’après
Ross “litteris pulcherrimis”, sans doute IV® s.).

Zußapıs. Sur un fragment de tuile, “graffito on the glazed side, obviously written on the
sherd”, début du VIE s., D 6, Pl. 7. En lettres sinistroverses, on lit: T’opyias ho Zvßapıos”.
Probablement ici un Gorgias fils de Sybaris, comme l'entend l'éditeur, avec un métronyme
("mother's name"), ce qui fournirait un nouvel emploi ancien de ce procédé?^, plutót

qu’avec un ethnique irrégulier de Sybaris®. On sait, en effet, qu’il a existé un nom de
femme Zvßapıs, et que Thémistocle l'avait donné à l'une de ses filles, les autres étant
denommees Aoia et Tradia, suivant un passage de Plutarque, Thémist. 3276. L'exemple
épigraphique de l'Agora serait à ajouter au dossier.

TpBar(A)os. Sur une base de lekane, graffito partiellement lisible, du deuxième quart du
V? s., F 62, PI. 13. Il doit s'agir de l'ethnique des Triballes, écrit sans gémination du lambda,
déjà attesté comme nom d'esclave?" ; ici, le statut du signataire est impossible à déterminer.

Tupoavos. Sur la base d'un lécythe, début du V* s., F 44, Pl. 12. On a ici un exemple
ancien et intéressant de l'ethnique des Tyrsénes, employé comme nom d'homme; Bechtel
ne pouvait citer que le féminin plus récent Tuppnvic?? , et l'on a aussi Tuppnvos, par ex-
emple SEG XXVII, 735 (II° s. de notre ère). On notera ici la graphie non attique Tupoa-??.

Paioorwc. Sur un skyphos à vernis noir, du second quart du V* s., F 77, Pl. 14. Légende
au génitif de possession: bargoriov, avec redoublement graphique du sigma”. C’est donc
un exemple ancien de l’ethnique de Phaistos, utilisé comme nom. Il ne figure pas encore
chez Bechtel, qui mentionnait baioros, forme identique au toponyme“! , ainsi que le dérivé
Paworwvvas particulier à la Crète*?. Mais pour le nom du père du célèbre Epiménide, la
tradition hésite entre Patorwos et Paioros®. La signature de l’Agora montre que la pre-
mière forme est parfaitement plausible.

Notes

! Mabel Lang, The Athenian Agora, Vol. XXI, Graffiti and Dipinti, Princeton, 1976. Pour des raisons

typographiques, la graphie des formes attiques a été modernisée ici en ce qui concerne les voyelles
longues (emploi de éta et oméga).

? Document repris chez O. Masson, Les inscriptions chypriotes syllabiques, seconde édition augmentée,
1983, no. 369b, fig. 154. Pour les signes du syllabaire commun, ibid., fig. 1—2.

Seu Masson, ibid. 413, index, et surtout dans Kypriakai Spoudai 28 (1964) 8—12 (liste qui serait à
compléter).

Pour les noms en ‘EAAo-, voir déjà F. Bechtel, Die historischen Personennamen des Griechischen. . .
(ci-aprés HPN), 1917 (réimpr. 1982) 152, et mes remarques dans Etrennes de septantaine, Mélanges
Michel Lejeune, Paris, 1978, 126-127.

5 L. Robert, Hellenica II, 1946, 75.

S er Rolley et O. Masson, BCH 95 (1971) 295—304; Inscr. chypr. syll, seconde édition, p. 422,

no. 369a, fig. 154.

Strack, Antike Münzen Nordgriechenlands II.1, 1, 1912, 54, no. 44; J.F. May, The Coinage of

Abdera, 1966, 138 sq., nos 184—185 (2 exemplaires).

8 Voir les listes abondantes chez R. Miinsterberg, Die Beamtennamen auf den griech. Miinzen, ré-
impr. 1973, 22-26, etc.

? Dans Glotta 32 (1953) 219-222 (= Kl. Schriften, 1959, 241-243), Manu Leumann a étudié divers
noms en ve, comme ®pdovs, '"Avópuc, etc. Le premier correspond à l'adjectif 0paov«, d'autres sont
clairement des diminutifs, cf. la liste de Bechtel, HPN, 52, au sujet de 'Ap8 pvc.

Leumann, Homerische Wörter, 1950, 141 sqq.; Chantraine, Dict. étymol. de la langue grecque, s.v.

102 Je l'étudie dans Rev. Phil. 1986.

A Déjà dans Athenian Agora XII, 1970, no. 627, PI. 27; méme texte, reproduit chez M. L. Lazzarini,
Le formule delle dediche votive nella Grecia arcaica (Memorie Lincei, 1976), 224, no. 346 (non
repris dans le SEG).

Bechtel, HPN, 348 sq. (existence de composés en -opnros) n’est nullement assurée).

3 Rev. Phil. 1963, 220 (à propos de -uoproc et Mopro-).

“i

BIRRA

BRE S

SREB

Ai
42
43

259

Liste chez Bechtel, ibid. 156—159. La combinaison d'éléments qui sont employés dans l'onomasti-
que d'une méme famille permet d'expliquer la formation de noms nouveaux, d'aspect souvent “illo-
gique" et irrationnel.

Cf. L. Threatte, The Grammar of Attic Inscriptions, Ll, 1980, 136. On trouve cependant trois exem-
ples d"Oraco« dans la Prosopopographia Attica.

Bechtel, HPN, 614; du méme, Attische Frauennamen, 1902, 134.

H. Solin, Die griech. Personennamen in Rom, Ein Namenbuch, 1982, 1224-1225.

O. Masson, "Vocabulaire grec et anthroponymie. . .", Mélanges offerts à P. Chantraine, Paris, 1972,
119--122.

Par exemple chez H. Wuthnow, Die semit. Menschennamen. .. , Leipzig, 1930, 70; pour l’Egypte,
dans les recueils de Preisigke et Foraboschi.

Wuthnow, ibid. Pour les noms MLK et MLK’ a Palmyre, voir J.G. Stark, Personal Names in Palmyrene
Inscriptions, Oxford, 1971, 95.

On remarquera en passant qu'un féminin assez ancien Maixa attesté en Attique, IG 112, 9112, aussi
12028, repr&sente un nom bien grec, Bechtel, HPN, 494, avec une interpretation assez plausible.
Chez Bechtel, HPN, 339, un exemple de Clazoménes (IV* s.).

Texte reproduit dans SEG XX, 735b I, 1. 115 et 116 (nominatif et génitif).

D.M. Robinson, AJA 51 (1947) 368, Pl. 89b-c; cf. J. et J. Robert, Bull. Epigr. 1949, no. 61; non
repris dans le SEG.

Article cité, avec renvoi à Bechtel, Att. Frauennamen, 7-8 (liste des composés normaux en -apiorn).
Voir L. Threatte, o.c. 121 et 391; Chantraine, Dict. étymol. s.v. Yäuabos.

L. Threatte, o.c. 400, avec des paralléles.

Bechtel, HPN, 592—595.

Voir aussi M.J. Milne et D. von Bothmer, Hesperia, 22 (1953), 220, pp. 66b, avec SEG XIII, 32
(commentaire).

Bechtel, HPN 543 et 546.

L. Threatte, o.c. 486 (devant beta), avec aussi cet exemple.

Bechtel, HPN, 597; Att. Frauennamen, 106. Davantage de détails dans Spitznamen, 1898, 76.

Mabel Lang corrige ainsi une lecture ‘‘XiBaxéo” donnée d'abord dans Hesperia Suppl. VIII (1949)
407.

Je me suis occupé à plusieurs reprises des métronymes: voir en dernier lieu BCH (1983) 397.
Hypothèse formulée par D.M. Lewis dans SEG XXIX, 287.

Texte utilisé par Bechtel, HPN, 551 et 554; Att. Frauennamen, 60-62.

Bechtel, HPN, 543, avec IG 112, 1951, 23 (rectifier ainsi la référence dans Agora XXI, 34). L. Ro-
bert, CRAI 1968, 419, signale aussi un To@aXXdc xonorde dans IG 112, 12822, ainsi qu’un homme
libre, père d’un mercenaire à Abou Simbel, A. Bernand, O. Masson, REG 1957, 30, n. 22, à propos
d’un TpogaAnds militaire en Bactriane, Robert ibid. 418--420 (milieu du III s.).

Bechtel, HPN, 546.

Voir sur ce point Threatte, o.c. 536.

Threatte, o.c. 529 (type ‘Apwoaros, il n’y a pas lieu de supprimer le second sigma).

Bechtel, HPN, 553.

Ibid. 549.

M. Guarducci, Inscr. Creticae I, 1935, 47, avec les auteurs concernés; aussi les articles correspondants
du Pape-Benselcr.

On terminera par une note sur le graffite C 23, Pl. 5. Dans ce texte obscene (avec le composé aristo-
phanesque Aakkómpcokroc), un nom “Suspduaxos” est non seulement “inconnu”, mais impossible.
Comme l’a remarqué D.M. Lewis, SEG XXIX, 287, il convient d'envisager un lapsus et de retrouver
(K)vspduaxos (la main est trés hésitante).

Some Remarks on Yasna 34

M. A. Mehendale (Bhandarkar Oriental Research Institute, Pune)

1. Yasna 34 2c: It has the expression pairigaeQe x$mavató vahme which has been rendered
by Humbach (1959) as ‘bei der die Herden übermittelnden Verherrlichung von euresglei-
chen’. On the strength of certain parallel passages Humbach believes that pairigaeOa- stands
for something like *pairi.dà.gaéOa- or *pairi.dàta.gaeQa- meaning 'den lebenden Besitz über-
mittelnd’. Insler (1975), on the other hand, feels that pairigae&a- means ‘universal’. He
compares with this word the Vedic expression párijman- which, according to him, means
‘around the earth, universal’. He therefore concludes that pairígaeOa-, which literally means
‘around the creatures’, also means ‘universal’. He translates the above expression as ‘in uni-
versal glory (p. 55) or praise (p. 221) of your kind’.

The context in which the above expression occurs is of a glorification (vahma-), accom-
panied by praise-songs, performed by an individual, and yet, apparently, many persons are
concerned in this act of glorification as is shown by the plural form ahma ‘by us’ in the
preceding stanza.’ Hence it seems better to interpret pairigaeOa- as surrounding the living
persons in the household’, i.e. (a glorification) in which all the members of the household
participate.

2. Yasna 34.5ab: kat vò xSaOram ka i$tit YyaoOanai mazda yaOa va ahmi?, aXà vohü
mananha Orayoidyai drigum yusmakam. Insler’s rendering runs as: ““Have ye the mastery,
have ye the power, Wise One, for the act to protect your needy dependent — as I indeed
am — with truth and with good thinking? " He thus takes the parenthetic phrase yaOa va
ahmi with reference to drigum yüXmakoam. He connects aa voRü managhá with Orayoidydi,
for he does not approve of treating it as commitative instrumental with vocative mazda.

It seems, however, preferable to relate the parenthetic phrase with those words between
which the phrase stands, viz. §yaoQanai and a¥a vohu manayha, and not with those which
stand away from it. This also means that one connects aša vohu manayha with XyaoOanai.
What is meant is that Zarathustra wants to know what kind of might, what kind of power
Ahura Mazda has for (1) the performer who acts (lit. an act performed) in accordance with
truth, and with good mind, as indeed, Zarathushtra himself is (yaQda vā ahmi), i.e. a per-
former of a truthful act done with good mind, and (2) for protecting those who are depen-
dent on Ahura Mazda.

3. Yasna 34.9ab: yöi spantam ärmaitim ... duS.SyaoOana avazazat. Humbach and Insler,
following Bartholomae, treat duX.$yaoOana as an adjective and render it, respectively, as
‘die Ubeltiter’ and ‘those of evil actions’ (nominative plural masculine). But it is better to

262

treat it as a noun ‘evil action’ and render the line as ‘those who have abandoned virtuous
piety by evil action’. This interpretation is recommended, on the one hand, by the close
proximity of duë.*yaoQana and avazazat, and, on the other, by the fact that duX *yaoOana
is thus contrasted with vanhauS manayho SyaoQana ‘with (good) act stemming from good
thinking’ of stanza 10.

Since stanzas 9 and 10 are related in that the activities of the bad and the good are con-
trasted, it is possible to imagine du¥.xratu-, as opposed to huxratu- (st. 10), as the subject
of 9 ab: yoi (7 duX.xratavo) spantam armaitim . . . du$.}yaoOanà avazazat.

4. Y. 349c: aéibyo ma aa syazdat. Humbach treats aša as nominative plural and trans-
lates, ‘von denen sollen sich schnell die Wahrhaftigkeiten entfernen’. Insler, however, treats
asd as instrumental singular and, accordingly, translates the passage as ‘from them one has
certainly retreated in accordance with truth’. It is, however, worthwhile to consider if
armaiti can be the subject: ‘(those of evil intentions who have abandoned piety), from
them (piety) shall quickly retreat along with truth’. In stanza 10 piety is said to be hiOgm
ašahyā ‘Genossin der Wahrhaftigkeit”? , and hence it is reasonable to suppose that when she
recedes from the perpetrators of evil actions, she does so along with truth. When evil-
thinkers abandon piety, piety too, along with truth, leaves them.

5. Yasna 34.13ab: tam advanam ahurä yôm moi mraoë vanhouS manayhô, daenä saoXyantam
ya hu.karata afacit urvaxSat “die Bahn, o Lebensherr, die du mir als die des Guten Gedan-
kens genannt hast, die gutgebahnte auf der die Sinne der Kraftspender mit Wahrhaftigkeit
wandeln” (Humbach). Humbach thus treats daend as the subject and hu.karata as referring
to advan.

Insler, however, reads yahu karata and treats karata as nominative singular masculine of
karatar- ‘extoller, commemorator’ and daénd as accusative plural feminine to which the
relative yahu refers. He translates: “To that, Lord, which Thou hast told me to be the road
of good thinking, to the conceptions of those who shall save, along which Thy extroller
shall proceed in alliance with truth indeed. . ."

With karatar ‘extoller’ Insler compares carakaraQra- ‘hymn of praise’ (Y. 29.8c). It seems
possible to connect these two words etymologically with Vedic kiri- (kirin- ?)^.

Notes

1 Also cf. dāmā (plural) in stanza 3, and gaëOù vispa in the same stanza which is nominative plural

according to Insler.

Humbach prefers to read kahmi ‘wie wenn ich schlafe’.

Insler, however, interprets this as referring to Ahura Madza.

On the doubtful nature of the stem kirin-, cf. M.A. Mehendale, BSOAS 37.1974:670—671.

References

Humbach, Helmut, 1959. Die Gäthäs des Zarathustra, Band I, II, (Heidelberg: Carl Winter).
Insler, S., 1975. The Gathas of Zarathustra. (Leiden: E.J. Brill).

Folk-linguistics and the Greek word

Anna Morpurgo Davies (Somerville College, Oxford)

1. Debts should be acknowledged even when they cannot be repaid. I have endless other
obligations but there is no-one to whom I am more deeply indebted, both personally and
intellectually, than Henry Hoenigswald. He has encouraged me when I hesitated, helped me
when I was in trouble, corrected me when I was wrong, and above all rescued me from
obscurity when I was confused. He has been friendly when we agreed, friendlier when we
disagreed. This is not easy to express without pomposity and sentimentality — two sins of
which Henry could never be guilty. I risk incurring his disapproval but, on this occasion at
least, KaAos 6 Kivöuvos.

In 1966 Hoenigswald pointed out that if the model of social sciences is to be taken at all
seriously linguists ought to be interested “not only in (a) what goes on (language), but also
in (b) how people react to what goes on (. . .) and in (c) what people say goes on (talk con-
cerning language)”. He also warned that “it will not do to dismiss these secondary and
tertiary modes of conduct merely as sources of error" (Hoenigswald 1966: 20). Folk-lin-
guistics, the discipline whose study Hoenigswald advocated, is closely linked with ethno-
graphy of speech and ethnolinguistics but is more decidedly slanted towards a study of the
speakers’ reactions to language and of their talk about language. It is an obvious hypothesis
that it has a direct and unique link with the history of linguistic thought or, as we more
normally say, the history of linguistics. In what follows I want to start from a specific
point made by Hoenigswald in his 1966 paper and examine it in some detail with reference
to Greek and to the beginnings of Greek linguistics.

2. Hoenigswald (1966: 18) considered the advantages of a more technical investigation for
those cultures in which we observe the existence of “a formalized activity in which lan-
guage is the object of reverence, contemplation, or study", and noted that the first ap-
proach to the subject “is necessarily by way of vocabulary"; we might start by asking, for
instance, for the equivalents of items like *word' or 'sentence' and inquire, inter alia,
whether they exist and are used in a definable sense. In this context he referred to “the old
observation that the Greeks had no word for *word""', but since his purpose was of a more
general nature, the observation remains isolated in the article. In my turn I want to ask
what notion, if any; the Greeks had of ‘word’. My question is not of a linguistic nature
(what was a Greek word? how do we define a Greek word?); it is rather a metalinguistic
question, if I can use ‘metalinguistic’ in this extended sense. Yet, it is not, or not pre-
dominantly, a question in the history of linguistics; it is mainly a question in Greek folk-
linguistics.

264

Why is the question worth asking and answering? First, because an answer will form an
integral part of an account of the Greek secondary responses to language. Secondly, be-
cause, if the answer is based on evidence which is not entirely lexical, it may lead to some
more general observations about the parallelism — or lack of it — between metalinguistic
lexicon and metalinguistic reactions. Thirdly, because it is at least possible that the notion
of word belongs both to folk-linguistics and to linguistic analysis proprie dicta. If so, an
answer — or the work leading to it — may serve to test the existence and nature of the link
and/or continuity which I hypothesized earlier between folk-linguistics and linguistics. If
this link exists it is in Greece that it ought to be most apparent. The whole tradition of
linguistic thought in the West at least as far as the nineteenth century is directly or indirect-
ly based on Greek thought; no doubt at each stage cultural influences of varying types were
also important but they did not destroy the basic continuity. From a Western point of view
it is in Greece that scholarly awareness of language starts; if so, it is in Greece that we
ought to test the connection between folk-linguistics and scholarly work. There are other
ways in which one could justify both the question and the time and labour spent in an-
swering it; for the moment I see no advantage in going into greater detail.

3. What evidence do we have at our disposal? Obviously we must start with the lexicon,
but two other types of evidence spring to mind. First, writing and all possible forms of
script; secondly, explicit statements which we may be lucky enough to have from any of
the preserved texts.

The lexicon has been explored already and Hoenigswald's statement quoted earlier is fre-
quently found in the secondary literature and is basically correct (Fournier 1946: 225;
Allen 1974: 113, note 1). In Homer éros and uddoc are frequently translated with ‘word(s)’,
but both terms would be more correctly rendered with ‘speech’, though uddos at least may
also refer to thought or story. We have no evidence that Homer had the lexical resources
necessary to ask questions such as ‘is X one or two words?’ or ‘are these two words of the
same type?’'. In Herodotus we find that éros can refer to one word (Békos in the Psam-
meticus story of 2,2, the name ‘Aouay in 2,30), but also means proverb (e.g. TÒ raAatóv
Eros in 7.51) or verse (e.g. Ounpov éros in 4,29) and can be used to refer to a sentence or
a phrase (e.g. ëri dé éme . .. in 3,82). Of the other possible terms, uùdos is very rare and
means “story, myth”; pqua is not frequent, occurs mostly in the plural and is best translated
with ‘saying(s)’ or ‘speech’; Aóyoc has its usual wide range of meanings. In Attic the closest
equivalents to ‘word’ are Aóyoc, priua, and ôvoua. It is hardly necessary to reiterate that
Aóyoc (which eventually became the technical term for 'sentence") means far more than
“word”; bmua is not much better defined. In Plato it refers to a proverb or saying (e.g.
Prot. 343 b), to a phrase or expression such as Au vidos (Crat. 399 a), to a single word
(Tim. 49 e), to a predicate or verb (Soph. 262 a). The modern translator is constantly in
doubt: should the ovundeias pnuarwv Te Kalbvoudrwv mentioned in Theaet. 168 b be ren-
dered as “sens coutumier des expressions et des mots” (Dies 1955: 196), “habitual use of
expressions and words" (McDowell 1973: 42), or — less likely — does it refer to verbs and
nouns as implied by LSJ s.v. dvoya (vi.2)? Similar problems arise for óvoua. In Homer it
refers only to ‘fame’ or to ‘proper name’. In Attic it may mean ‘word’, but also ‘name’ and,
later, ‘noun’ as contrasted with bqua ‘verb’ or ‘predicate’. In the famous passage from the
Cratylos mentioned earlier (399 a) does Plato speak of the shift from the phrase Au yidoc
to the name AápiAoc or of the shift from a phrase to a single word ??

265

Later on further terminology — which is now technical terminology — comes in. Dionysius
of Halicarnassus (First Century B.C.) uses óvoua for word in his earlier works but in his
later contributions speaks of ovduara or Aé£eic or uôpia Adyou (Schenkenveld 1983: 70).
Householder (1981: 4) points out that Apollonius Dyscolus (Second Century A.D.) uses
three equivalents for ‘word’: uópiov, AdEts, and pépos Adyou or uépoc TOU Aöyov; Övona,
however, still appears as a cover-all term (Ap. Dysc. de synt. 18, 2). We may discuss the
exact value of these forms but there is little doubt that by this stage the one-or-two-words
and the word-type questions I hypothesized earlier can be asked and answered. I shall
return to this later.

What emerges from this brief survey is not, as has been stated, that the notion of word
appears late in the development of Greek thought, but rather that the Greek lexicon does
not offer any clear evidence for the existence of such a notion; yet, it does not offer any
evidence for its absence either. Even when technical thinking about grammar starts and a
notion of word is obviously available, the vocabulary does not make this immediately clear.

4. The easiest way of defining a word in English is with reference to the written language:
a word is a sequence of letters written between empty spaces and/or some form of punc-
tuation. This statement is obviously trivial and, in any case, can only refer to word-forms
rather than, say, lexemes, but may well account for the confidence with which the layman
speaks of words, in spite of the notorious difficulties encountered by linguists in their
attempt to produce a definition of this concept.* A parallel statement is impossible for
classical Greek. In the absence of spacing, accents, and smooth breathings, there was no
formal device used to distinguish words in the written language. Admittedly one letter (A)
in some regions at least occurred only in word-initial position and some letters could not
occur in word-final position, but here writing reflects speech. We do not know that these
facts were the object of metalinguistic observation and consequently we do not know
whether they are relevant to our inquiry. If so, do we learn anything from the written lan-
guage?

The answer is definitely positive if we take into account the whole gamut of writing which
we find in ancient Greece. Of the three scripts used in Greece for Greek (Linear B, syllabic
Cyprian, and the alphabet) it is only the third, and even this in a limited part of its history,
which seems to provide no evidence for the Greek notion of word. The first two definitely
repay further analysis and the alphabet itself is not as barren of information as at first ap-
pears to be.

The data vary in time and space. Linear B was written in the second half of the Second
Millennium B.C. in Crete, in the Peloponnese, at Thebes, and possibly in other parts of My-
cenaean Greece. Syllabic Cyprian, with the exception of one early document, dates from
the first part of the first Millennium B.C. and stretches until the Hellenistic period; most of
the texts come from Cyprus? I do not need to rehearse here the basic data about the
Greek alphabet. The three scripts are independent of each other. There is genetic affinity
between Cyprian and Linear B, but we do not know the exact nature of the relationship.
There is a chronological overlap between Cyprian and the alphabet but until the latest
period at least no obvious influences of one on the other.

If what we aim at is a view of what the ancient Greeks in general or, at any rate, the man in
the ancient Greek street felt about language, we ought to use the evidence of writing with

266

extreme caution. The level of literacy in Greece was low at all times and for the potential
users of Linear B and syllabic Cyprian must have been very low. It is only in the fifth
century and only in some parts of Greece that writing acquired a different status and im-
portance. Yet all that we can hope to reconstruct is the metalinguistic assumptions and
beliefs of a small subset of Greeks; if so, we should not be deterred by the nature of our
evidence. That the number of literate speakers was not unbearably small even in the earlier
period is shown by the Linear B texts of Knossos. It seems probable that these all belonged
to the same year and were written by a minimum of one hundred different scribal hands.
The odds are that the scribes were not a caste separated from the others; there must have
been a number of administrators who could write but had other tasks as well (Olivier
1977a; 1977b).

The syllabic scripts are often compared unfavourably with the alphabet. The latter is
praised for its quasi-phonemic character, its simplicity and clarity — qualities which are
missing in the former. Linear B and, to a lesser extent, Cyprian are reproached for the
ambiguity of their graphic systems and their incapacity to express all grammatical distinc-
tions. Yet in our strictures we ignore a basic point: the syllabic scripts provide information
which the alphabet does not offer. From the alphabet, for instance, we learn very little, if
anything, about syllabic division; from the syllabic scripts we learn a great deal. For a ques-
tion which is closely related to that from which we started — that of word division, i.e.
written identification of word-forms — the syllabaries are highly informative and far more
obviously so than the alphabet.

5. Let us start from Linear B. The script has a sign used as a word-divider which has no
exact match in any standardized form of the Greek alphabet until spacing is introduced in
the Middle Ages. The regular use of the word-divider is an innovation of Linear B; the sign
is not found — or found only sporadically -- in the earlier Cretan scripts to which Linear B
is related.

In Linear B the word-divider is normally a small vertical bar (the size varies) and is used

with some regularity between sequences of syllabograms to separate them one from the

other. It does not occur at the beginning of a text nor does it normally occur at the beginning

or end of a line — which effectively prevents us from speaking of a preposed or postponed

‘word-determinative’. The word-divider can also — but need not — be omitted

(a) when it is replaced by an empty space,

(b) when the size of the characters changes (e.g. from large to small), between a word and
the following one,

For the rest in the vast majority of the cases we find a word-divider where a modern editor

would leave spacing between words. Yet there are exceptions which can be subsumed

under a few main types and ought to be considered.

First, the word-divider may be occasionally and unpredictably omitted, though none of the
points made above apply. The omission seems accidental because the same sequence is
written by the same scribe with and without word-divider (here conventionally indicated
with a comma); cf. KN Fp 1, 10 (hand 138) a-ne-mo, i-je-re-ja vs. Fp 13,3 (hand 138)
a-ne-mo-i-je-re-ja (twice); PY Fn 187,8 (hand 2) u-po-jo-po-ti-ni-ja (the editors doubt the
existence of a word-divider) vs. Fr 1236 (hand 2) u-po-jo, po-ti-ni-ja (cf. also Fr 1225,1);
PY Aq 64,2,4,14,15 to-to, we-to vs. ibid. 6,7,16 to-to-we-to.

267

Secondly, Linear B apparently avoids marking with the word-divider graphic words which
would consist of one sign only; hence the spelling to-so-pa: tos(sjos pas in PY Ja749 and
Jn601,9 contrasted e.g. with the plural to-so, pa-te (KN B 1055,9). An isolated sign could
be taken for an ideogram or an abbreviation, so that presumably there is a graphic motiva-
tion for this rule.

A third set of exceptions could conceivably belong to our second category but is best
treated separately. This consists of a number of particles, pronouns, etc. which are regularly
not divided from the word which precedes or follows. Thus -de and -qe (Gr. ôé, Te) are
always joined to the preceding word (and followed by the word-divider), jo- and o- (Gr.
cos 227) are joined to the word which follows. Not all these spellings can be explained
merely on the basis of the Mycenaean reluctance to write one-sign words separately since
we find two-sign sequences treated in the same manner: da-mo-de-mi : damos de min (PY
Ep 704,5), e-ke-de-mi : ekhei de min (PY Na 926), probably o-da-a; and o-de-qa-a; ; the
negative particle o-u- (Gr. où) is always joined to the word which follows : o-u-di-do-si,
o-u-di-do-to, o-u-pa-ro-ke-ne-[, o-u-te-mi, o-u-wo-ze (o-wo-ze)$. The form o-u-ki (Gr. obxi)
occurs only once and is joined to -fe-mi which follows (KN V 280,5; cf. o-u-te-mi ibid.,
11-14). By contrast o-u-ge (Gr. oùre) is frequently found at Knossos and at Pylos as an
independent word, preceded and followed by the word-divider. No doubt we deal here
with elements comparable with the prepositive and postpositive elements of the metricians
(Maas 1962: 84ff.; West 1982: 25f) but the problem is how our particles are to be defined.
I wonder whether a merely distributional and syntactical account is sufficient. Mycenaean
has a number of disyllabic prepositions (o-pi, e-pi, pa-ro etc.); in the noun phrases in which
they occur they are normally (not always) separated from the noun which follows by a
word-divider. Distributionally everything points to a ‘prepositive’ status of the preposition;
yet the ‘prepositive’ o-u ‘not’ is regularly joined to the following word; the prepositions are
not. The different spelling may be explained if pa-ro was orthotonic, but o- was not. For
later Greek it is often suggested that the prepositions were proclitic; in Mycenaean the posi-
tion must have been different, but the suggestion agrees with all that we know about the
historical development of the prepositions, which certainly started as orthotonic and
presumably nominal elements.”

A preliminary conclusion at this stage is that in Mycenaean the word-divider is used to
separate accentual groups, i.e. speech sequences characterized by one main accent; the ex-
ceptions are explained by graphic reasons. However, before this can be accepted, we must
consider two further sets of exceptions.

I have mentioned above the odd instances of omission of word-divider due presumably to
negligence or sometimes to lack of space. There are also two instances of sequences which
ought to be written with a word-divider but occur consistently and more than once with-
out. In Knossos pa-si-te-o-i : pansi theoihi ‘to all gods’ occurs 15 different times in tablets
written by four different scribes (Fp-, Ga-, Gg- passim); in Pylos we-te-i-we-te-i ‘every year’
(Feret-Ferec) occurs 13 times in the same tablet (Es644) again without word-divider. In the
second sequence we may well have an instance of univerbation. The formation is archaic? ,
we have no other examples of it in Mycenaean, and it may have survived in a fossilized and
‘univerbated’ status with consequent accent simplification.? The case of pa-si-te-o-i is more
puzzling. Semantically, the odds are that ‘All gods’ count as a unit, a single denomination;
yet, we expect that if we had more texts we would also find a nominative *pa-te(-)te-o :
pantes theoi and a genitive *pa-to te-o : panton theon with both elements declined, so that

268

on grammatical grounds it would be difficult to attribute word status to the phrase. The
position is only partially comparable to that of the Homeric kapn Kopowvres for which
even oi madatot, according to Eust. 165, 14f., disputed whether it was one or two words
(cf. also Schol. in Hom. Il. B 11 b Erbse). In the absence of further evidence we must con-
clude with a non liquet, though I doubt that this is sufficient counterevidence to destroy
the suggestion made earlier.

Another difficulty concerns the odd instances of compounds in which the first element is
separated from the second by a word-divider. These are constantly quoted in the literature
(most recently Risch 1983: 377£.), but no attempt is normally made to assess the extent
and the implications of the phenomenon. In fact, in constrast with the very large number
of compounds attested in the Linear B tablets and written as a single word, there are only a
minimum of three and a maximum of five compounds with this abnormal spelling. They
are all attested in a single set of tablets, the Ta tablets of Pylos: ke-re-si-jo, we-ke : Kresio-
werges 'of Cretan workmanship' (Ta641,1.1; Ta 709, 3.3); e-ne-wo, pe-za : en(njewopeza
‘nine-footed’ (Ta 642, 1.3; contrast e-ne-wo-pe-za Ta 713, 1.3; Ta 715, 1); pu-ko-so, e-ke-e :
pukso-{h)ekhehe “with box-wood support’ (?) (Ta 715,3); a-pu, ke-ka-u-me-no | : apukekau-
menos ‘burnt off? (Ta 641, 1); a-pi, to-ni-jo 27? (Ta 716, 1). The first three forms must be
compounds because the morphology of the second element would otherwise be unex-
plainable. The position of the last two words is more difficult to understand: a-pi to-ni-jo
is obscure and though it could be a compound there is no certainty about it. In a-pu ke-ka-
u-me-no there is just a chance that a-pu is adverbial, though otherwise tmesis is not found
in Mycenaean. It would be rash to base any conclusion on these two forms, but the first
three compounds are disturbing: if the word-divider is used to mark separate accentual
units, and if these words are compounds, i.e. — in Greek terms — accentual units, why are
their elements separated by the word-divider? Two suggestions are possible. First, we could
argue that besides the accentual and other criteria which justify the treatment of com-
pounds as single words, the Mycenaean scribes consciously or unconsciously made use of
semantic criteria which induced them to treat compounds as formed of separate words.
The problem is not normally discussed but I suspect this is the view which would be
generally accepted. Yet, it does not explain why the divided compounds are so limited in
number and limited to one set of tablets. A second possibility, and one which I favour, is
that the scribe of the Ta tablets produced new words while he was writing (‘nine-footed’,
‘made-in-Crete’, ‘with-box-wood-support’) and that he marked off the separate elements in
order to facilitate comprehension — much in the same way in which British English uses
hyphenated forms (‘tax-free’) or open forms (‘reading material’) to indicate newly formed
compounds which have not yet gained permanent status and consequently are not written
as solid words (‘bedroom’). "!

For Mycenaean the conclusion must be that the use of the word-divider is mainly deter-
mined by accentual criteria, though conceivably other criteria may play some part. We may
feel entitled to argue that the Mycenaean scribes had a notion of word (word-form) con-
sciously or subconsciously defined in accentual terms; from an English point of view this
would mean that ‘blackbird’, ‘cat’, ‘the cat’, and ‘kill him’ are all treated as one-word se-
quences. On the other hand it would be wrong to generalize from here and argue that the
Mycenean scribes analysed the spoken chain in predominantly phonological terms. It seems
likely, and a few pieces of evidence confirm it, that the syllabification pattern in most cases
ignored word-boundaries; thus if the syllabification was expressed in writing in the usual

269

manner for a Mycenaean phrase [ona:ton (h)ekhei] we would expect a Linear B spelling
*o-na-to-he-ke. Yet what is really attested is o-na-to, e-ke (PY En-, Ep-passim) or even
o-na-to-e-ke with omitted word-divider and omission in writing of -n in word final position
(PY Ep 539, 3; 613,7): correct word division is more important than correct syllabifica-
tion.!?

6. Such much about Mycenaean. What about the other scripts? In the Cyprian syllabary
word-dividers (bars, dots, empty spaces) exist (Masson 1972: 107ff; 1983: 68ff.) but are
not used with as much regularity as in Mycenaean and sometimes mark phrases rather than
individual words. Yet, word division is more reliably indicated by another device. In Linear
B the consonants which close a word are normally not written; in Cyprian the final con-
sonant of a word is written with a sign of the Ce type: cf. pa-si-le-u-se : BaoıXevs; po-to-li-
ne : nroAw. In internal position different rules apply: cf. ka-a-si-ti : ypaodı; pa-ta : ravra.
It is noticeable that here too word-division takes priority over syllabification: in the se-
quence m7oAw "ESaXwv the final -n of m7oAw must have formed a syllable with the first
vowel of 'E6aXov but the script renders it as if it belonged to mroAw only: po-to-li-ne-e-ta-
li-o-ne, not *po-to-li-ne-ta-li-o-ne.

If we can judge from the spelling, monosyllabic enclitic and proclitic elements are not
treated as independent words'?: cf. a-u-ta-ra-mi (not *a-u-ta-re mi) for abrap ui (ue);
ta-sa-ke (not (ta-se ke) for raç ke. Other enclitic particles are mat, vv etc. Some proclitics
are easily identified: cf. to-no-ro-ko-ne: rov ópkov (not *to-ne o-ro-ko-ne); su-tu-ka-i: avv
ruxa (not *su-ne tu-ka-i). Yet, as Hermann (1906: 240ff.) saw a long time ago, the distinc-
tion here is between particles which end in a sibilant and are written as independent words,
and particles which end in a nasal and are joined to the word which follows; contrast ta-se
a-ta-na-se : tas Adavas (Masson 1983: 217, 20) and to-i-se Ka-si-ke-ne-to-i-se : ro kaor
qverois (ibid., 5) with ta-po-to-li-ne : rav nroÂw (ibid., 1) and ta-na-ta-na-ne : rav 'A9avav
(ibid., 27). In spite of the different renderings, Hermann assumed that all these elements
(particles, articles etc.) were proclitic, and explained spellings such as ta-se : ras, po-se :
nos, ka-se : kas ‘and’ as due to some form of hypercorrection or linguistic conservatism.
Intervocalic -s-, he agreed, had changed into -h-, a sound impossible to indicate with the
normal graphic resources of the syllabary; the spelling of the ta-se type was meant to support
the etymologically correct pronunciation with -s. The speakers had no difficulty in uttering
a sibilant in word-final position or before pause and consequently that type of spelling was
adopted. The texts published after Hermann’s article have not changed the position. His
explanation may be correct and is accepted by Masson (1983: 70) but a difficulty remains.
For e.g. kas we have before a word which began with a vowel two possible spellings from a
relatively early date: either the Ka-se o-na-si-lo-i : Kas ‘Ovaotdor type or the Ka-a-ti : ka(c)
avr, type. The latter example shows indeed that -s either dropped or became - thus
supporting Hermann, but Hermann has not succeeded in explaining why the third possible
(and expected) spelling *ka-so-na-si-lo-i is normally not attested’ , though all early inscrip-
tions preserve intervocalic -s- (in writing at least). As an alternative one might suggest that
in word-final position -s tended to become -h and consequently the -se- sign was felt to
have both an -s- and an -h value, differently from e.g. -sa-, which was not used to express
final -s. An article like tas may have been pronounced [tah] and this may have called for
one of two spellings, either ta-se or ta. If so, the Cyprian writers may have sacrificed the
correct rendering of syllabic and word junction to a correct rendering of the consonantal

270

nature. This explanation too cannot be demonstrated, and the problem remains open, but
we must agree with Hermann that the odd spellings of the fa-se, ka-se type must be due in
some way to the presence of the s; they do not prove that these words were not proclitic
or were not felt as such.

I now return to the original question: what categories did the Cyprians mark in writing?
The use of the word-divider is too irregular to be of much use; the use of the -Ce conven-
tion is more informative. It shows that a ‘word’ was either an orthotonic word or a se-
quence of orthotonic and unaccented elements.!* The unexpected spelling of e.g. Tas as
contrasted with e.g. rav may imply that a new concept of word was evolving (perhaps
under the pressure of phonological facts) and that the proclitics were attributed a more
independent status than the enclitics.’° The evidence is not compelling and it may be safer
to retain from this discussion one conclusion only: accentual factors played a considerable
part in determining the notion of word which was influential in Cyprian spelling.

7. Cyprian is important in that it may provide us with the link we need between the Linear
B and the alphabetic texts. Mycenaean writing, as we have seen, reveals an implicit notion
of word mostly based on accentual criteria; we ought to ask whether this notion survived
the collapse of Mycenaean civilization. Cyprian shows that, admittedly in a relatively iso-
lated area, this happened. Should we now assume that those Greeks who wrote in the
alphabetic script well after the Mycenaean period but contemporaneously with most of the
Cyprian texts shared this notion?

In classical Greece most inscriptions do not mark word division and smooth breathings and
accents are not written. Yet some data are available. Some texts avoid breaking words at
the end ofa line. Others use bars or dots or other devices as a form of interpunction within
the sentence or paragraph. This happens in a few regions only and mostly in the sixth and
fifth centuries B.C., though earlier and later exampies are found; within this limited period
we can gain some information.!”

With the notable exception of Threatte's (1980: 73ff.) detailed discussion of Attic data
until the third century A.D., there is no modern study of the punctuated texts, and most
editions of these inscriptions are unreliable in their rendering of punctuation. What
follows is simply a first attempt at making use of a material not sufficiently analysed.

We cannot assign a single purpose to punctuation. We find it at the end of a paragraph or a
line or a verse; it may also be used to mark the first word of a paragraph or an important
name or number; sometimes it may even serve as a decorative element. In some texts it
seems to divide words, in others phrases and/or words. The extraordinarily rare cases where
it occurs in the middle of a word must be due to a mistake!? ; we could go further and say
that one of the important points is that punctuation does not occur in the middle of a
compound or a word. In a few texts punctuation is clearly connected with word division in
a predictable manner. Dots, bars etc. correspond to a high degree with the spacing intro-
duced by the modern editor, but in addition to a few unpredictable divergences there are
some regular disagreements. We may use as evidence two sets of examples: first, the Teiae
Dirae, a fifth century imprecation from Teos written on two stelae, of which one is known
only from old drawings and editions (the stone was lost) and one has been recently
found?" ; secondly some archaic texts from Argolis: a representative sample is provided by
LH Jeffery (1961: 151ff., plates 26-31). Finally, a countercheck is possible thanks to
Threatte’s analysis of the archaic texts from Athens.

271

The Teiae Dirae make regular use of the colon, two dots one above the other. This is
omitted at the end of the line, but otherwise separates words. À more complete statement
would further add that: |

1. no form of the article is separated from what follows (76, rdu, THV, 76, TA, TOS, TAS) and
the prepositions are treated in a similar manner (èk, èv, ert, ës, Kara, mept, npôs, bnô(?));
2. the particles % ‘or’ and kai ‘and’ are not followed by the colon and the same applies to
the negatives ob, obôé, un;

3. two of the previously listed elements may follow each other (conjunction * article, pre-
position + article) without any intermediary punctuation or punctuation after the second
of them;

4. two postpositives (ôé and ay) are attested and again they are not preceded by punctua-
tion.?!

The archaic inscriptions of Argos and Mycenae vary in their use of punctuation (normally
dots). Some do not have it at all, one (Jeffery 1961: 169 no. 20) has a series of oddities
(punctuation after ras, whole phrases marked by punctuation), others (Jeffery 1961: 168
nos. 6, 8, 9, p. 169 nos. 19, 22, 26, 28, p. 170, no. 32; p. 174, no. 2) divide words according
to criteria very similar to those found in the Teiae Dirae: the article in its various case
forms and the prepositions are not followed by punctuation nor are ai ‘if’, kai ‘and’, and
ut) ‘not’; the postpositives 5é and re are not preceded by punctuation.”

For Attic Threatte (1980: 79ff.) has listed the main categories to be considered. Inter-
puncts characterize private inscriptions of the sixth and fifth centuries but are rarer in
public texts of the fifth century, where they also have more specialized uses; in the texts
where the main purpose is that of dividing words the usual rules apply: prepositions, kat,
and the forms of the article are not separated from the word which follows.”

Thus the general pattern of written ‘word-division’ is relatively consistent; barring mistakes
and oddities, a ‘word’ is either an orthotonic word or a sequence of proclitics — orthotonic
word — enclitics. Expertise in word-division, thus defined, was obviously more necessary
for the Mycenaean or Cyprian scribes than for their alphabetic counterparts, but it is
remarkable that all three shared similar conceptions of the ‘word’. Continuity of school
between Mycenaean and alphabetic writing is probably to be excluded; if so, we ought
to reach the conclusion that the Mycenaean, Cyprian and alphabetic writers based their
principles of word-division on a common response to speech which consciously or more
probably unconsciously, analysed it on the basis of accentual criteria. The continuity
between the three traditions is also important for another reason: it allows us to say that
for the literate part of the population a notion of word, or, more exactly, of word-form,
was available at a much earlier stage than the terminology might induce us to believe. We
are now in a position to argue on the basis of some concrete evidence against the assump-
tion that the metalinguistic lexicon provides a full account of metalinguistic responses.”

8. In addition to lexicon and writing I mentioned above a third type of evidence, that
provided by the explicit statements of writers, thinkers and scholars. Here we are on the
borderline between folk-linguistics and linguistic analysis; a distinction cannot always be
made but a brief inquiry may be relevant to the problem I raised earlier about the con-
tinuity or otherwise between these two forms of response to language. What follows is by
necessity allusive, incomplete and superficial. A fuller statement would call for a complete

272

account of what we know and do not know about the development of philosophical and
linguistic thought in Greece.

A first observation is obvious. The recognition of ‘eight parts of speech’ which is character-
istic of traditional grammar presuppose some notion of word or, more exactly, lexeme,
since in effect the ‘parts of speech’ are word-classes. The standard histories of linguistics
connect the eight parts of speech with the Alexandrinian grammarians and regularly refer
to Dionysius Thrax. It is not surprising to see that Dionysius’ Techne also offers a defini-
tion of word (Ae£ıs) as the smallest syntactical element of sentence structure (Adyos).”° If
so, and in spite of the justified complaints by the scholia about the vagueness of Dionysius’
definition, it is reasonable to assume that an explicit notion of word was available to
Dionysius and his contemporaries — and probably to his predecessors. Yet, the increasing
doubts about the authenticity of the Techne” make it difficult to base any chronological
statement on this evidence so that we must turn to other sources. The definition of Aé£uc
as qoi) eyypáuuaroc which Diogenes Laertius (VII 56) attributes to Diogenes of Babylon
is ambiguous as to the meaning of Aé£i (word or speech with emphasis on sound rather
than sense?).”’ I shall discuss later better attested grammarians like Apollonius Dyscolus.

What about the earlier period? Here the evidence is difficult to assess and I can only men-
tion a few relevant facts. The study of Homeric or poetic words must have started before
the fifth century. Aristophanes (fr. 222 Koch) refers to them with the term yAwrrat which
becomes or had become a technical term for rare or special words (cf. Arist. Rhet. 1410 b,
Poet. 1457 b, see Cassio 1977: 75ff; for Greek Latte, 1925: 136ff; cf. also Pfeiffer 1968:
12, 41, 78f).

It may be argued that this scholarly interest in some types of words belongs more to lin-
guistics than to folk-linguistics, but the interest in etymologies (word-etymologies), which
is even earlier, straddles between the two fields. Later on, Prodicus' concern with synonyms
and the discussion about dpdörns dvouarwv reflected in Plato's Cratylus (to give just a few
instances) show an interest in language which focuses on words, their semantic analysis and
their etymology. Yet, this is more relevant to a study of the general attitudes to language
than of attitudes to words as such. The literary sources do not tell us what counted as a
word, with what criteria, if any, speech was segmented, and whether the distinction be-
tween word-form and lexeme was at all clear; nor do we know at what stage questions of
‘wordhood’ were first raised. For Aristotle (and for Plato??) óvoua and pua, ‘noun’ and
‘verb’, may be word-classes (lexeme classes), but in chapter 20 of the Poetics (if authentic)
Aristotle lists under the heading of uépr rc Aé£ecoc ‘parts of speech’ a mixture of word-
classes (noun, verb, etc.) and other units below and above the level of the word (orotxetov,
syllable, sentence etc.) This makes it unlikely that at this stage the notion of part of
speech presupposes the notion of word, as it does in the later grammarians and probably in
the Stoics. On the other hand the introduction of the notion of mraot — however ill
defined or ill understood — implies that in some instances Aristotle distinguished word-
forms from lexemes.? Later grammarians will make this explicit distinguishing between
the parts of speech (word-classes) and their maperöueva/accidentia (inflection etc.). A
concern with the category ‘word’ also appears in Aristotle's statements about compounds,
which he clearly takes as single words. Just as Plato had contrasted the pr)ua. "Ad qoc?
‘dear to Zeus’ with the óvoua (name! or Sword") ‘Apos’ (Crat. 399 a), Aristotle (de int.
16 a) contrasts the Aóyoc 'kaAóc inmos’ ‘beautiful horse’ with the dvoua 'KáAAurmOC and

273

enunciates a rule which he then generalizes. The -irnos element of KaAXırrroc has no mean-
ing of its own; the individual elements of compounds differ from those of simplicia in that
they have some force, but nevertheless are not significant in isolation (cf. also poet. 1457 a
and see Ackrill 1963: 115ff.; Belardi 1976: 117ff.). In these statements there is a fair
amount of unclarity; nor is it possible to see how far the formal divergencies between the
compounds and the phrases with which they are contrasted matter and how far the defini-
tion is purely semantic. However, there is little doubt that the question of ‘wordhood’ has
been raised.

So far we are still remote from seeing any link between the evidence provided by the
various forms of writing we have examined and that provided by the literary sources. On
the one hand we found some techniques for segmentation which seemed to presuppose a
notion of word largely based on accentual criteria; on the other hand we have frequent
citations of words, semantic and etymological discussions, etc., but we hear little about
accentual criteria. Admittedly, if the text is not too corrupt, Aristotle is likely to have
mentioned the oVvdeopor among his parts of speech (poet. 1457 a) and these certainly in-
cluded enclitic forms. Yet since, as we have seen, not ail Aristotelian parts of speech are
word-classes this need not be significant; in other words, we do not need to assume that for
Aristotle an unaccented element could be a word. The later grammarians will also list
ovvSeopo and äpŸpa among their parts of speech (word-classes) and there the implication
must definitely be that some unaccented forms counted as words, in contradiction with the
notion of word which we extrapolated from a study of writing. Yet before reaching the
later stage we find sufficient evidence to prove that the accentual notion of word had some
influence on Greek thought. We can look again at the famous passage from Plato's Cratylos
which I have quoted earlier (399 a); we are told that in order to make of the pnua Au
yihos an dvoua we must remove the second iota and pronounce the middle syllable grave
rather than acute (Aui poç > Aptos). This does not certainly amount to an accentual
definition of word but shows conscious knowledge of a link between accentual factors and
univerbation. In a passage of the Sophist (237 c) Plato refers to roùvopa ... roro, TO ut) v
‘this name, not-being”.?? The use of övoua to refer to two words would be exceptional; it
seems easier to assume that for Plato un öv was a single word. A similar point, again in
connection with a negative particle linked to the following word, may be made à propos of
Aristotle's discussion in de int. 16a,b of phrases like obk ávOpconoc, obx byıaveı. These, we
are told, are not nouns (6voya) or verbs (priua), nor is there a special name with which to
call them; hence they can be called óvoua àópirov or pra àópwrov. I do not want to
discuss the philosophical or logical implications of this view, but it seems plausible that
Aristotle would not have found it necessary to discuss the point or at least to discuss it in
the way he does, if he had not conceived of obk Gvdpunos and oby bytaver as single
words.?! Here too the suggestion is that the proclitic negative is treated as part of the word
which follows.

Finally I turn very briefly to some later and more detailed discussions of words which we
find in the grammarians. I shall consider two examples only, both from Apollonius Dysco-
lus? Syntax. In III 123ff. (262 b) Apollonius begins to discuss the subjunctive and in 132
he tackles the problem raised by the conjunctions which normally precede this mood. He
wants to argue that e.g. 6páumc can occur by itself and not only as part of e.g. àv 6pdunc.
Yet he is not too certain of his ground and the discussion is protracted (ibid. 133): “But if
this needs further proof, we will add a little more to show that conjunctions are not

274

compounded with verbs into single words (eis Ev kepos Adyov). [If they were] how could
other words intervene? E.g. éà» orjuepov kat aüpwov axovons ...”. (Translation by House-
holder 1981: 203). Here Apollonius appeals, in the first instance at least, to the principle
which Bloomfield much later will call the principle of indivisibility, but he is stopped in
full flight by an (anachronistic) objection for which he himself is responsible. In Homer, we
are told, an expected participle xareó78 coc can be replaced by Kara ... &öndwc with an-
other word intervening between the preposition and the verbal stem. Yet we still want to
argue that the form is a compound; if so, we must admit that divisibility is possible in
compounds though they are single words. Does it follow that édv Spduns is a compound?
Apollonius now invokes another principle (also used by Bloomfield in his discussion of
words), that of structural regularity. The phrase édv dpduns has the same structure as
phrases like iva Spdunc or öppa Temo(9nc, but here the two conjunctions ipa and öppa
have accents of their own, so that they cannot be parts of a compound. Accordingly, éav
dpayns must also be treated as a phrase and not as a compound.

There is a third principle here and must be treated as conclusive: two main accents cannot
belong to the same word. The progression of the argument seems clear. Starting point is the
assumption that táv belongs with the following word, an assumption which, I would argue,
was part of the general belief in the earlier period. This is rejected for the reasons indicated.
Apollonius seems to have found arguments to support his conclusion: while two main ac-
cents must belong to two words, one main accent is not the mark of a single word. Proclitics
and enclitics are unaccented and yet they can count as independent words.”

A question which concerns the nature of words is discussed later in the same work (IV
26ff., 318 b ff.) with reference first to duri and Kaddrt, which Apollonius does not want
to treat as single words, and then to verbs of the kara'ypáwco type, which for Apollonius
are compounds and not phrases. Apollonius exploits the same principles as before with
some additions or modifications. He mentions phonological facts (in Greek h can only
occur in wordinitial position) (IV 28), morphological data (the genitive Neac IlóAecoe
shows that Ne HO cannot be a compound) (IV 34), structural patterning (IV 41), and
finally he offers a more extensive discussion of the accentual principle (IV 47). The in-
dividual elements of a compound may have the same accentual pattern in composition as
when they are used independently: cf. IlepiuxAurds vs. KAuTOS. In this case observation of
the accent contributes little to our understanding of the grammatical status of the sequence.
Yet we also find forms like ovvoida contrasted with the simplex oda; in this case the ac-
centual difference clearly indicates that ovvotéa is a compound. The level of sophistication
reached at this point is quite high; the obvious comparison is with the standard contrast
drawn in modern textbooks between the English word blackbird with the main stress on
the first element of the compound and the phrase black bird with the main stress on the
second word.

For our purposes the interesting result is that in Apollonius, as no doubt in all his predeces-
sors, there is still hesitation in the identification and definition of words. Part at least of
these hesitations must have been determined by the conflict between an earlier, more
primitive view based on the ‘one accent, one word’ principle which I have argued was the
current ‘folk-linguistic’ view, and the more sophisticated attitudes developed in the course
of philosophical and grammatical discussion. In the Greek tradition continuity is not
surprising. At a later date — when we cannot say — the anonymous author of the de
prosodiis, a pamphlet preserved together with the Techne in some manuscripts and at-

275

tributed to Dionysius, refers to odx oürws, i.e. to a phrase formed by a negative particle
and an adverb, as to uia Aé£u ‘one word’. The scholiast, a more sophisticated scholar,
reproaches the author for this mistake: oùy odrws is two words, not one (Gramm. Gr. I 1,
p. 112-3 Uhlig). Obviously the discussion continued. Another scholiast in his commentary
to the Techne produced a fuller definition of word which anticipates much of the discus-
sion of the twentieth century: “a word is the smallest linguistic sound sequence (www
eyypáuuaroc), which is indivisible, can be said by itself, understood by itself, and is
pronounced with one accent and one breathing(?)" (Schol. Marc. in Gramm. Gr. 1 3,p. 352
Hilgard). We find here the antecedents of more than one modern definition: the word as
an accentual unit, the word as an indivisible unit, the word as a minimum free form, the
word as a unit of meaning, etc. In this form the scholiast’s definition is eclectic and un-
workable (though no more so than most of this century's definitions); yet it gives witness
of the long and lively discussions which led to it, and perhaps of the ‘pre-scientific’ ‘folk-
linguistic’ period which preceded these discussions.

9. The general conclusion is brief. During most of its history the ancient Greek lexicon
offers no exact equivalent for our ‘word’, i.e. it has no word with a comparable range of
meanings. I have argued that some notion of word was nevertheless available and found its
expression in the various forms of writing used in Greece from the Second Millenium on-
wards. The exact nature of the notion is of course difficult to establish but it is likely that
the scribes or stonecutters identified words mainly on accentual criteria: one main accent
— one word. The literary texts are ambiguous and often uninformative, not so much be-
cause they do not confirm our belief that some concept of word was available (that they
do), but because they give us little, if any, information about the nature of this concept.
Yet, Plato, Aristotle, and, much later, the grammarians show signs of a deeply ingrained
view of the word as accentually identified. Philosophical and grammatical thought chal-
lenged or refined this view but I suspect that we ought not to underestimate its importance.
In this particular case at least it may be legitimate to argue for some form of continuity
between folk-linguistics and linguistics.”

Notes
! For the Homeric words rendered with ‘word’ cf. Fournier 1946: 211ff.; cf. also Hofmann (1922). In
Homer Aówo« (in the plural) is attested only twice with the meaning ‘words, speech’; the semantic
‘expansion’ of Adyoc probably happened later.

In Aristotle, where the word closest in meaning to ‘word’ is övoua, Bonitz (Index Arist. s.v. Adyos
I 2) recognizes three passages where Aóyoc = ‘word’. In two of these (Nich. Eth. 1103 b; Soph. El.
169a) the meaning is ambiguous (e.g. evi 67) Aóyc may be translated with ‘in one word’ but in fact
refers to a sentence or a phrase); in Rhet. 1406 a Adyos seems to refer to one word, a compound, but
could well mean something like ‘expression’.

It is also worthwhile to look at the metalinguistic vocabulary of the Derveni Papyrus which must
count as the earliest literary commentary we have in Greece (see for a provisional edition ZPE 1982
and cf. West 1983: 75ff.; Burkert 1985): &mos means ‘verse’, Adyos seems to have a generic value,
bvoua means ‘name’, pia probably means ‘expression’ and once (col. xxii 8) refers to one word. Of
Demetra it is said that Anunrn[o 86] cvouác6n Wonelp] ATH Mirnp, ek Guyotépwr Elv] óvoua
(col. xviii 9—10).

For a classical discussion of the notion of word, cf. Bloomfield (1983: sections 1.5 to 11.7). For
more modern references and discussions ci. e.g. Allen (1973: 22ff.), Matthews (1974: 20ff. and
159ff.), and the two book-iength accounts by Kramsky (1969) and Pulgram (1970). An elementary

276

but very clear discussion is that by Palmer (1971: 41ff.) who reaches no positive results: “In con-
clusion, sadiy, we have to say that the word is not a clearly definable linguistic unit” (1971: 51).
The oldest Greek document from Cyprus is now a syllabic inscription of the eleventh century B.C.
with one Greek name in the genitive; cf. Karageorghis (1980: 122ff.). For the Cyprian texts in
general cf. Masson (1983).

Exact references are not always given here but can be easily traced through Olivier-Godart (1973),
Ventris and Chadwick (1973: 527ff.) and, for the first part of the alphabet, through Aura Jorro
(1985).

For the Mycenaean prepositions see Morpurgo Davies (1983), where, however, the accentuation is
not discussed. — I have not mentioned above Myc. dwo ‘two’ which seems to contradict the rule ac-
cording to which graphic words composed of one sign only are not written in isolation. This spelling
occurs in Pylos only and the dwo sign is probably an invention of a Pylos scribe; since it is formed by
two wo signs facing each other it may well be that the sign was felt to have a different graphic status
from that of e.g. the simple sign pa used to write the word for ‘all’ in the phrase to-so-pa.

Cf. Masson (1966), Dressler (1968) and more recently Dunkel (1981: 214ff. with note 2). The inter-
pretation of KN a-mo-ra-ma as an iterative ämorämar (cf. Gr. fiuap ‘day’) is too doubtful to be used
in this context.

Dunkel (1981) has argued that for Indo-European we can only reconstruct iterative compounds
based on preverbs. If so, the we-te-i-we-te-i formation cannot be inherited and one cannot quote the
accentuation of Vedic divédive etc. as a parallel. Nevertheless composition and accent simplification
may have occurred independently in Vedic and Greek. Moreover the Cyprian a-ma-ti-a-ma-ti 'every
day’ discussed by Masson (1966) is also written as one word in an archaic inscription (seventh or
sixth century) where word-division indicated by a dot seems to be relatively regular.

I ignore ka-ka re-a of KN R 1815 (written over erasure). A possible explanation is that by Risch
(1983: 384) but no word divider is present and it is possible that what the scribe had in mind was
*ka-ka-a-re-a (cf. for the formation Ruijgh 1983: 395f.).

Cf. e.g. Quirk and Greenbaum (1973: 1019). — Though the compounds were ee treated as
single words it is also true that the Mycenaean scribes ‘responded’ to compounds in a different way
from that in which they responded to simplicia. This was demonstrated (though with different pur-
poses) by Ernst Risch (1983: 374ff.) who succeeded in showing that the scribes adopted special
spellings (sometimes prompted by etymological considerations) in order to mark off the different
elements of a compound. The importance of Risch’s work for a study of Greek attitudes to language
is very great.

The odd mistake is attested: cf. te-ko-to-na-pe (vs. te-ko-to-a-pe), i-qo-na-to-mo, ko-to-na-no-no; see
Risch (1983: 378). Obviously in addition to syliabification questions of juncture may also be rele-
vant.

I have deliberately spoken of monosyllabic proclitics and enclitics because the position of disyllabic
appositives, if they existed, is doubtful. The first person singular of the verb ‘to be’, e-mi, is normally
separated in writing from the word which precedes, but Masson (1981: 647, note 78) now quotes
one example of e-mi written in scriptio continua with the preceding word and discusses a Kafizin
(hence late instance of èr: written in the same manner.

A possible exception is provided by the recently published texts of Kafizin (Mitford 1980: nos.
118b, 133b, 191, 303) of the late third century B.C. As shown by Masson (1981: 648), they
regularly write e-se (és), ta-se (rac), and even ka-se (xas), but in the passages quoted above they have
ka-se-i-ko-so-to-i or the like for the expected kas eixo0To- instead of the required *ka-se-e-i-ko-so-to-i.
We are prevented from reading kas ixooro-, which would cause no orthographic problems, by the
spelling ka-e-i-ko-so-to-i: ka eıkoorwı of no. 252. It is difficult to understand what is happening, all
the more so in view of the ‘regular’ spelling of the other -s prepositives.

Compounds are regularly written as single words with the exception of e-xe o-ru-xe found four times
in the Idalion inscription (Masson 1983: no. 217, 12 [twice], 24, 25); there may be graphic or
phonetic reasons for this spelling (cf. Masson 1983: 241) but we do not understand them.

We should not forget that that of proclitic is a modern and not an ancient concept (cf. e.g. Schwyzer
1939: 350ff.). It is also noticeable that in Cyprian proclitics which end in a vowel call for special
syllabification rules (Masson 1983: 69); cf. to-sa-ta-si-wo-se: ré Eraow^oc, which implies that the
whole sf- cluster is assigned to the first syllable of the name and is not divided between the first and
the second syllables of the word-group. Linguistically this is not extraordinary. Morphological factors

277

can influence syllabification; cf. for instance, the Attic scansion of &x-Aecrco with along first syllable,
though normally clusters of muta cum liquida are assigned to the second syllable (Snell 1958;
Ancher 1978: 66f.). More interesting is the fact that the Cyprian spelling may reveal some awareness
of the phenomenon.

The origins of punctuation must be Semitic (cf. Larfeld 1914: 302ff.). For our purposes, however, a
discussion of the Semitic antecedents is not necessary since the linguistic response in which we are
interested is Greek and Greek only. Similarly, it did not seem worthwhile to investigate the use of
the ‘word-divider’ in Cretan scripts other than Linear B.

Still useful, in spite of its date, is the Berlin dissertation by Rudolf Kaiser (1887) with the earlier
literature. Some detailed comments are also available in Larfeld 1914: 302ff.; 1907: 564ff. and for a
brief discussion see also Guarducci 1969: 391ff. For a preliminary study I found essential the archaic
inscriptions collected in Jeffery (1961), where the various forms of interpunction used in archaic
Greece are discussed in detail.

See Kaiser (1887: 23ff.), Threatte (1980: 82—3). For some non-Attic examples cf. e.g. the seventh
century base from Naxos with p a:vedéxe (Schwyzer 1923: no. 757; Jeffery 1961: 304, no. 3, pl.
55) and the fifth century block from Ephesos with Sixafov:res (Schwyzer 1923: 462, no. 708,
Jeffery 1961: 344, no. 55b).

For the new stele cf. Herrmann (1981); for the old stele (now lost) see e.g. Schwyzer (1923: no.
710); Meiggs and Lewis (1969; no. 30) and for a full list of the old editions cf. Herrmann (1981:
1—2, notes 2—8).

The importance of the Teiae Dirae for a study of interpunction was underlined by Kaiser (1887:
17—19) who drew the right conclusions: “In universum igitur statuere licet, eam esse rationem, ut
omnes illae voculae encliticae atque procliticae interpunctione a nominibus suis non seiungantur"
(cf. also Larfeld 1914: 303). For the punctuation it is customary to refer to Jacobsohn (1909: 107
note 1), but Jacobsohn has far less to say than Kaiser (1887) or Larfeld (1914). — One word is
necessary about the standard editions of the old stele. Interpuncts are sometimes omitted in Meiggs-
Lewis (1969: no. 30; cf. B 9, 29, 32, 39 twice) and in Schwyzer (1923: no. 710; cf. A 1, B 31, 32);
Schwyzer also adds an interpunct in B 8. Since the stele is lost it is now necessary to return to
Boeckh CIG 3044 and Roehl IGA vi. 12. In the new stele Herrmann’s edition (1981) shows no inter-
puncts in a) 21;b) 10, 11;c) 4, possibly d) 12, d) 22. The photographs make it likely that there
were interpuncts in c) 4, d) 22 and possibly in b) 10. The other passages are too broken to be checked,
with the possible exception of d) 12 where doric sé rıuoxewv does not seem to be punctuated after
6é — nor am I certain that we would expect it to be.

There are two interesting sets of exceptions in Jeffery 1961: 168 no. 8 (= SEG XI 314, XXII 263).
First, in a list of demiourgoi which consists of six names joined by 7e kai interpuncts regularly
precede re and follow kai (all instances of re are at the end of a line, so that there may have been
‘potential’ interpuncts after re too). This contrasts with the rest of the inscription, where kat occurs
twice without being followed by an interpunct and re occurs once immediately after a noun (with
no interpunct) and before interpunct + «at. The odds are that the interpuncts are meant to emphasize
the names of the demiourgoi. Secondly, the dative plural of the article roto. appears three times
before an interpunct (and a noun or another word). It is not sufficient to explain this with the
disyllabic form of the article, since the disyllabic preposition eni is not followed by an interpunct.
It is conceivable that this form was orthotonic as indicated by the conventional accentuation.
Threatte (1980: 23) observes that “in a number of texts no real rationality seems to lie behind the
placing of the interpuncts”. This is occasionally so, but there is no reason to list among these texts,
as Threatte does, IGAA p. 143, no. 57 = SEG 10.461 (= Hansen 1983: no. 27). The text is punctuated
as one would expect for word-division; the only problem is that one colon is missing in the phrase
napa céua Bavdvros.

So far I have deliberately refrained from mentioning the Near Eastern evidence for word-division,
since the question of possible influences is obviously insoluble and in any case is not likely to be
important (cf. above note 17). Yet in most cuneiform writings the use of determinatives and logo-
grams implies some concept of word; in cuneiform Hittite there is the added factor that the Glossen-
keil marks off a word rather than a phrase. Akkadian does not normally separate words in writing
but Hittite leaves empty spaces between words and significantly for our purposes treats as one word
a sequence of an orthotonic element and any number of enclitic particles. In the First Millennium
B.C. (the Second Millennium data are too obscure) the Hieroglyphic Luwian script separates words

278

with a special sign which in some texts is used with absolute regularity. There too the enclitic parti-
cles are not separated from the orthotonic element which precedes them. Finally, we should not
forget the extensive grammatical work done in the Near Fast at a much earlier date than in Greece
(and in India). The peculiar linguistic situation created a need for glossaries and these presupposed
some notion of word. Also the Old Babylonian grammatical texts of the Second Millennium B.C.
give evidence for the setting up of paradigms which presumably presupposed both an analysis into
word-forms and some notion of lexeme (cf. Jacobsen 1974: 41ff.).

Gramm. Gr. 1 1 p. 22 Uhlig: Ad£ıs tori uépos eAdxioTOvr TOD KaTa GUVTa~W Adyov. I have followed
Robins’ translation (1957: 91); a more literal version is that by Arens (1969: 23): “Das Wort ist der
kleinste Teil des auf Zusammenfügung beruhenden Satz”.

For an introduction to the question see Pinborg (1975: 69ff. and especially 103ff.). A list of selected
recent additions to his bibliography ought to include at least Siebenborn (1976, with the review by
Fehling 1979), Erbse (1980), Ax (1982). For the parts of speech cf. also Wouters (1979: especially
179ff.) and Schenkenveld (1983). For the outsider the main problem is that the whole chronology
of traditional grammar is now under discussion and that it is difficult to know whether the old view
according to which grammatical analysis had reached its traditional form well before the second-
first centuries B.C. is still correct.

The odds are that here Ae&ıs does not mean word or only word; cf. Pinborg (1975: 97) and see also
the old observations by Schmidt (1839: 18ff. with notes 31, 32, 37); cf. also the German translation
by Hülser (Schmidt 1979: 107, note 31 a). i

There is a further problem here since Aristotle uses rrwox not only for inflection etc. but also for
what we would call derivation: in Rket. 1410 a the adjective xaAkovs, ‘made of bronze’ and the
genitive xaAkoò of xaAkdc, ‘bronze’ are treated as nrwoeıs of the same Yvoua. We would say that
xaAkobc and xaAkóc are different words or lexemes. The literature about Aristotle’s nroous is
immense; for a survey see Calboli (1972: 86ff.), but cf. also Belardi (1979: 11ff.).

I am grateful for this reference to my colleague Lesley Brown.

For a general discussion cf. Ackrill (1963: 117ff.) and above all Belardi (1979) who rightly disagrees
with the views previously expressed by Pagliaro (1954: 29ff.).

This would be difficult to argue if, on the basis of the controversial passage Soph. El. 166 b 3 (cf.
poet. 61323), we had to assume that Aristotle treated the où negative as oxytone (see e.g. Gudeman’s
commentary to the relevant passage of the Poetics). Yet I fail to see why translators and commenta-
tors ignore Wackernagel’s arguments (1955: 1077ff.) in favour of an ‘oxytone’ od and a ‘baritone’
ov; in the absence of a proper refutation of what seems to me the only correct interpretation I see no
need to be worried by these passages and I am ready to argue that for Aristotle où was unaccented
(cf. Erbse's note in his edition of the scholia to Hom. 7. 23, 238). I do not think that this view must
be rejected on the strength of Herodian’s statement (1 494, 504 Lenz) that the ob negative dévvera:
like 1804 and tov (cf. Allen 1973: 253) since the grammarians always spoke of the proclitics as
oxytone (Vendryes 1945: 63ff. and 73). The observation that roùro è (scil. ov) kail ev rh ovvereta
oó£Uverac (Herod. I 504) is difficult to understand and cannot be given too much weight particularly
since ovveneca is a modern correction for ovim0eta of the manuscripts.

Obviously the whole of my paraphrasis and summary is coached in terms which are far too modern.
I must repeat again (cf. note 15) that ‘proclitic’ is a modern term (but for the notion of proclisis cf.
the observation by Quintilian 1 5 27 quoted by Allen 1973: 25: “‘cum dico circum litora, tanquam
unum enuntio dissimulata distinctione, itaque tamquam in una voce una est acuta”).

Aééts Fort EAaxio T quovr] €yypáuguaroc aueprs i6ía pntù Kai bla vonrn, bp'&va Tovov Kal nveüna
ayouern. I have hesitated in translating www eyypäuuaros, the phrase used by Diogenes of Babylon
in his definition of Adéts (cf. p. 272 above and note 26; cf. also [Plato] Def. 414 d) but it seems
likely that by this stage eyypduparos need not mean literally ‘which is written’ or ‘which can be
written’ but is closer to ‘articulate’ or ‘linguistic’; for an earlier period see already Apoll. Dysc. synt.
3, 180 where obk eyypappartwv < pwvwv > refers to meaningless sounds or noises.

This article was completed in 1983; except for a very few items it has been impossible to introduce
additions and alterations after that date.

279

References

Ackrill, J.L., 1963. Aristotle's Categories and De Interpretatione, translated with notes and glossary
(Oxford: Clarendon Press).

Allen, W.S., 1973. Accent and Rhythm (Cambridge: Cambridge University Press).

Allen, W.S., 1974. Vox Graeca (2nd edition) (Cambridge: Cambridge University Press).

Ancher, G.P., 1978. *Coupe morphologique et coupe syllabique en attique d’après l'emploi de la cor-
reptio attica dans le théâtre d’Eschyle”, RPh 52: 66-86.

Arens, Hans, 1969. Sprachwissenschaft (2nd edition) (Freiburg/München: Alber).

Aura Jorro, Francisco, 1985. Diccionario micénico 1 (Madrid: Istituto Bibliogräfico Hispanico).

Ax, Wolfram, 1982. “Aristarch und die Grammatik”, Glotta 60: 96-109.

Belardi, Walter, 1976. “Il semplice e complesso nella teoresi aristotelica della forma linguistica”, RCCM
18: 117-141.

Belardi, Walter, 1979. “Ptosis e onoma aoriston”, in: Studi su Varrone, sulla retorica, storiografia e
poesia latina. Scritti in onore di Benedetto Riposati (Rieti e Milano), 11—21.

Bloomfield, Leonard, 1933. Language (New York: Holt).

Burkert, Walter, 1985. “Der Autor von Derveni: Stesimbrotos Nepi Terer@wv”, ZPE 62: 1-5.

Calboli, Gualtiero, 1972. La linguistica moderna e il latino (Bologna: Pàtron).

Cassio, Albio, 1977. Aristofane, Banchettanti. I frammenti. (Pisa: Giardini).

Diès, Auguste, 1955. Platon, Théétète (3rd edition) (Paris: Les Belles Lettres).

Dressler, Wolfgang, 1968. ‘““Vedisch dive-dive und die indogermanischen Iterativ-Komposita”, in Studien
für W. Brandenstein (Innsbruck: Institut für vergleichende Sprachwissenschaft), 39-47.

Dunkel, George, 1981. ““Amredita and iteration of preverbs in Vedic and Hittite’, KZ 95: 214-226.

Erbse, Hartmut, 1980. “Zur normativen Grammatik der Alexandriner”, Glotta 58: 236—258.

Fehling, Detlev, 1979. Review of Siebenborn, E. Die Lehre von der Sprachrichtigkeit und ihrer Kriterien
(Amsterdam: Grüner), Gnomon 51: 488—490.

Fournier, H., 1946. Les verbes ‘dire’ en grec ancien (Paris).

Guarducci, Margherita, 1969. Epigrafia greca 1 (Roma: Istituto Poligrafico dello Stato).

Hansen, P.A., 1983. Carmina epigraphica graeca (Berlin: de Gruyter).

Hermann, Eduard, 1906. “Zur kyprischen Silbenschrift”, /F 19: 240-249.

Herrmann, P., 1981. “Teos und Abdera im 5. Jahrhundert v. Chr.”, Chiron 11: 1-30.

Hoenigswald, H.M., 1966. “A proposal for the study of Folk-Linguistics”, in: Sociolinguistics. Proceed-
ings of the UCLA Sociolinguistics Conference 1964, edited by W. Bright (The Hague: Mouton), 16—
26.

Hofmann, Erich, 1922. Qua ratione Eros, uöüßos, aivos, Adyos et vocabula ab eisdem stirpibus derivata in
antiquo graeco sermone (usque ad annum fere 400) adhibita sint (Göttingen: Dieterich).

Householder, Fred W., 1981. The Syntax of Apollonius Dyscolus (Studies in the History of Linguistics
23) (Amsterdam: Benjamins).

Jacobsen, Thorkild, 1909. “Very Ancient Texts: Babylonian Grammatical Texts", in Studies in the
History of Linguistics, edited by D. Hymes (Bloomington: Indiana University Press), 41-62.

Jacobson, H., 1909. “Beiträge zur Sprache und Verstechnik des homerischen Epos”, Hermes 44: 78-—
110.

Jeffery, L.H., 1961. Local Scripts of Archaic Greece (Oxford: Clarendon Press).

Kaiser, Rudolf, 1887. De inscriptionum graecarum interpunctione (Berlin: Otto Francke).

Karageorghis, Vassos, 1980. “Fouilles à l’Ancienne-Paphos de Chypre: les premiers colons grecs”,
Comptes-rendus de l'académie des inscriptions et belles lettres (Paris), 122-136.

Kramsky, J., 1969. The word as a linguistic unit (The Hague: Mouton).

Larfeld, Wilhelm, 1907. Handbuch der griechischen Epigraphik 1 u. 2 (Leipzig: Reisland).

Larfeld, Wilhelm, 1914. Griechische Epigraphik (3rd edition) (Miinchen: Beck).

Latte, Kurt, 1925. “Glossographica”, Philologus 80: 136-174 (= Kleine Schriften, München: Beck,
1968: 631-661).

Maas, Paul, 1962. Greek Metre (Oxford: Clarendon Press).

Masson, Olivier, 1966. “A propos de deux formules redoublées au locatif”, Ziva Antika 15: 257-266.

Masson, Olivier, 1972. “Remarques sur l’épigraphie chypriote syllabique: les procédés de séparation des
mots", in Praktika tou Prötou Diethnous Kyprologikou Synedriou (Leukösia) 1: 107-110.

Masson, Olivier, 1981. “A propos des inscriptions chypriotes de Kafizin”, BCH 105: 623—649.

280

Masson, Olivier, 1983. Inscriptions chypriotes syllabiques (2nd edition) (Paris: Boccard).

Matthews, P.H., 1974. Morphology. An Introduction to the Theory of Word-Structure (Cambridge:
Cambridge University Press).

McDowell, John, 1973. Plato, Theaetetus (Clarendon Plato Series) (Oxford: Clarendon Press).

Meiggs, Russell and D.M. Lewis, 1969. A Selection of Greek Historical Inscriptions (Oxford: Clarendon
Press).

Mitford, T.B., 1980. The Nymphaeum of Kafizin: The inscribed pottery (Berlin: de Gruyter).

Morpurgo Davies, Anna, 1983. "Mycenaean and Greek Prepositions: o-pi, e-pi etc.", in: Res Mycenaeae,
Akten des VII. Internationalen Mykenologischen Colloquium in Nürnberg, edited by A. Heubeck —
G. Neumann (Góttingen: Vandenhoeck & Ruprecht), 287 —310.

Morpurgo Davies, Anna, 1986. “Forms of Writing in the Ancient Mediterranean World”, in: The Written
Word (Wolfson College Lectures 1985), edited by G. Baumann (Oxford: Clarendon Press), 51—77.
Morpurgo Davies, Anna, 1987. “Mycenaean and Greek Syllabification”, to appear in: Tractata Mycenaea.

Proceedings of the 8th International Colloquium on Mycenaean Studies (Skopje).

Olivier, Jean-Pierre, 1977a. Les scribes de Cnossos (Roma: Edizioni dell'Ateneo).

Olivier, Jean-Pierre, 1977b. “Gli scribi di Cnosso”, in La Civiltà Micenea, edited by G. Maddoli (Roma:
Laterza), 39-44.

Olivier, Jean-Pierre / L. Godart et al., 1973. Index généraux du Linéaire B (Roma: Edizioni dell'Ateneo).

Pagliaro, Antonio, 1954. "II capitolo linguistico della 'Poetica' di Aristotele", RL 3: 1—55.

Palmer, F.R., 1971. Grammar (Harmondsworth: Penguin).

Pfeiffer, Rudolf, 1968. History of Classical Scholarship I (Oxford: Clarendon Press).

Pinborg, Jan, 1975. "Classical Antiquity: Greece", in: Current Trends in Linguistics 13/1, edited by T.
Sebeok (The Hague: Mouton), 69—126.

Pulgram, Ernst, 1970. Syllable, Word, Nexus, Cursus (The Hague: Mouton).

Quirk, Randolph/Sidney Greenbaum et all., 1973. A Grammar of Contemporary English (3rd impres-
sion) (Harlow: Longman).

Risch, Ernst, 1983. "Probleme bei der Schreibung von Hiat im Kompositionsfuge im Mykenischen”, in:
Res Mycenaeae. Akten des VII. Internationalen Mykenologischen Colloquium in Nürnberg, edited by
A. Heubeck /G. Neumann (Göttingen: Vandenhoeck & Ruprecht), 374-390.

Robins, R.H., 1957. “Dionysius Thrax and the Western Grammatical Tradition", TPS: 67—106.

Ruijgh, C.J., 1983. "Observations sur les neutres en -s/h-”, in: Res Mycenaeae, Akten des VII. Interna-
tionalen Mykenologischen Colloquium in Nürnberg, edited by A. Hcubeck/G. Neumann (Góttingen:
Vandenhoeck & Ruprecht), 391—407.

Schenkeveld, D.M., 1983. “Linguistic theories in the rhetorical work of Dionysius of Halicarnassus”,
Glotta 61: 67—94.

Schmidt, Rudolf, 1839. Stoicorum Grammatica (Halle: Anton) (reprinted Amsterdam: Hakkert 1967).

Schmidt, Rudolf, 1979. Die Grammatik der Stoiker (translation by H. Hülser of Schmidt 1839) (Braun-
schweig: Vieweg).

Schwyzer, Eduard, 1923. Dialectorum graecarum exempla epigraphica potiora (Leipzig: Hirzel).

Schwyzer, Eduard, 1939. Griechische Grammatik 1, (München: Beck).

Siebenborn, Elmar, 1976. Die Lehre von der Sprachrichtigkeit und ihren Kriterien (Amsterdam: Grüner).

Snell, Bruno, 1958. “muta cum liquida”, Glotta 37: 160.

Threatte, Leslie, 1980. The Grammar of Attic Inscriptions 1 (Berlin: de Gruyter).

Vendryes, Joseph, 1945. Traité d'accentuation grecque (2nd edition) (Paris: Klincksieck).

Ventris, Michael/John Chadwick, 1973. Documents in Mycenaean Greek (2nd edition) (Cambridge:
Cambridge University Press).

Wackernagel, Jacob, 1955. Kleine Schriften II (Göttingen: Vandenhoeck & Ruprecht).

West, M.L., 1982. Greek Metre (Oxford: Clarendon Press).

West, M.L., 1983. The Orphic Poems (Oxford: Clarendon Press).

Wouters, Alfons, 1979. The grammatical Papyri from Graeco-Roman Egypt (Bruxelles: Kon. Akademie
van Wetenschappen).

ZPE, 1982. “Der Orphische Papyrus von Derveni”, ZPE 47: 1—12 (after p. 300).

On Vowel Length in Gothic

William G. Moulton (Princeton University)

Because Gothic is the oldest Germanic language for which we have extensive data, the
study of its grammar — including its phonology — has long exerted a special fascination not
only for Germanists but also for Indo-Europeanists. I therefore hope that a discussion of
vowel length in Gothic is a fitting contribution to this volume written in honor of Henry
Hoenigswald — a great Indo-Europeanist who also has a lively interest in Germanic. I shall
begin with a review of the scholarship on my topic. This is a small but nevertheless inter-
esting chapter in the history of linguistics: as successive theoretical approaches have changed.
so also have interpretations of the Gothic vowels. In conclusion I shall try to make a con-
tribution of my own, using a theoretical approach which is by no means new but which has
thus far been largely overlooked.

Until the mid- 1950's it was taken for granted that vowel length was relevant in Gothic, just
as in the other older Germanic languages and in reconstructed Proto-Germanic and Proto-
Indo-European. A rather widely accepted phonetic interpretation of the graphic evidence
was as follows (Streitberg 1920):

Short. Graphic: a ui ai au
Phonetic: [a u i e >]

Long. Graphic: a ueie o ai au
Phonetic: [a:u: i: e: o e: o1]

Diphthongal: Graphic: ai au iu
Phonetic: fai au iu]

For ease of typing and printing, I shall symbolize vowel length by writing a colon
after a vowel symbol rather than a macron over it — regardless of the practice of the
authors whom I cite. This has the added advantage of reminding us visually that, in
terms of syllable structure, long vowels function as the equivalent of short vowel plus
consonant, i.e. /V:/ = /VC/. This is very astutely pointed out in Vennemann 1971:
106-7.
In order to symbolize the three assumed phonetic values of graphic ai, au, philologists
(following Jacob Grimm) customarily wrote them as follows: (1) short monophthongal
at, au, reflecting earlier short vowels, as in tathun ‘ten’, sathWan ‘to see’, bairan ‘to bear’,
dauhtar ‘daughter’, haurn ‘horn’; (2) diphthongal di, du, reflecting earlier diphthongs, as in
stdins ‘stone’, dugo ‘eye’; (3) long monophthongal ai, au, reflecting earlier long vowels
(before a following vowel), as in saian ‘to sow’, bauan ‘to build’. The strongly historical-
comparative nature of this three-way interpretation of graphic ai, au is obvious.

282

Where there was disagreement with this interpretation, it concerned not the assumption of
vowel length but rather the assumption of three different phonetic values each for graphic
ai, au. An example is Wright 1910. Throughout the text of his grammar he used the tradi-
tional interpretation and the following spellings: short a, u, i, a£, au; long a:, u:, ei, e:, 0:,
ai, au, diphthongal di, du, iu. But he was skeptical. At the end of his grammar he wrote: “It
seems almost incredible that a man like Ulfilas, who showed such skill in other respects,
should have used ai for a short open e, a long open æ: and a diphthong; and au for a short
open o, a long open 9: and a diphthong. Whereas, if we assume that the diphthongs had
become monophthongs (@:, 9:), there is nothing incongruous in his having used each of the
digraphs to represent two sounds which differed in quantity but not in quality" (362).
That is to say, he believed it more likely that Gothic had the following vowels (with minor
changes in symbolization):

Short. Graphic: a ui ai au
Phonetic: fa u i e 0]
Long. Graphic: a u eie o ai au
Phonetic: [a:u: i: e: o: e: 01]

Diphthongal: Graphic: iu
Phonetic: tiu]

Neither Wright nor any other philologist, however, doubted that Gothic showed a distinc-
tion between short and long vowels.

In the first "structuralist" treatment of Gothic phonology (Moulton 1948), I used for
vowels essentially the former of these two interpretations. Though my article was badly
flawed in its treatment of the vowels (in retrospect I have often wished that I had never
published it), it had at least one advantage: it stimulated a number of other structuralists
to write articles showing how wrong I had been. Reactions were not long in coming. In
1949 Bennett published an article on “The monophthongization of Gothic di du.” After
referring to scholars who accepted the traditional view that graphic di, au still represented
diphthongs at the time of Wulfila, he wrote: “In recent years, however, an increasing
number of linguists, including Hirt, Marstrander, Wright, and Mossé [references] have
reached the conclusion that Wulfila’s ai au were no longer diphthongs but had become
monophthongized to long open [e: 9:1 respectively” (15). He then gave arguments for this
view, which convinced me then and still convince me now. He did not, however, question
the relevance of vowel length.

The second reaction came in a 1950 article by Penzl, “Orthography and phonemes in
Wulfila’s Gothic.” He, too, assumed that traditional di, du had been monophthongized to
long /e:/, /o:/, but he also did something more important than this. In my 1948 article I
had assumed that some of Wulfila's spellings reflected allophones. Penzl not only showed
that I was mistaken; he also formulated a general principle. Speakers are not generally
aware of the allophones in their native language; hence only under very special conditions
can we expect them to symbolize allophones in writing. Applying this reasoning to Gothic,
Penzl wrote: “If Wulfila's orthography was created by a native speaker of Gothic and was
intended for other native speakers of Gothic, any representation of subphonemic variants
must seem somewhat surprising and demand a special explanation. . .” (217). This reasoning
convinced me; I should have known better. The final sentence in Penzl’s article reads as
follows: “It cannot be emphasized enough that Wulfila’s orthography and its relation to

283

the phonemic pattern must be the main consideration in any study of the phonemes of his
Gothic dialect” (230).

The implications of Penzl’s article are clear. Except in most unusual circumstances, the
orthography of a language will not indicate any subphonemic (allophonic) distinctions; it
will indicate only phonemic distinctions. To this there is a corollary. Can we also assume
that the orthography will indicate all phonemic distinctions; and, if a possible distinction is
not indicated in the orthography, is this proof that it did not exist? Four years later Penzl
(1954: 412) asked just this question: “Can we actually prove that [vowel] quantity is
phonemic in Wulfila's Gothic?" Though he raised this question, he did not attempt to
answer it.

An answer was not long in coming, namely in a 1955 article by one of Penzl’s students,
Marchand. After reviewing the evidence, Marchand concluded his article by writing: “It is
believed that a case could be made to the effect that [vowel] length was not phonemic in
Gothic, though the point cannot be pressed too far. . . If anyone states that length was or
was not phonemic in Gothic, he should realize that this is, to a great extent, an arbitrary
statement” (86). (Cf. also Marchand’s 1955 dissertation, later published in 1973.)

Marchand’s cautiously noncommittal statement concerning the relevance of vowel length
in Gothic was soon accepted more or less unquestioningly by a number of other struc-
turalists, notably Jones 1958 and Hamp 1958 (these were successive articles in volume 34
of Language) and Bennett 1964. As good structuralists, they then proceeded to describe
the full vowel “system” of Gothic, rejecting length as phonemic. This was done in the
greatest detail by Marchand (1958: 508):

Front Central Back
Close i (ei)

High Open I =

: Close e o
Mid Open e (ai) o (au)
Low a

The one remaining problem was how to integrate graphic iu into this system; various solu-
tions were proposed.

With the advent of a new approach to phonology, that of generative phonology, there came
a new series of works on Gothic; and, as we might expect, the generativists have been just
as fascinated by the phonology of Gothic as were the structuralists. Two of these works
have appeared only as dissertations, namely Buckalew 1964 and Beade 1971; and — inex-
cusably — I have not read them. (I know only, from Vennemann 1971 and Beck 1973, that
Buckalew rejected vowel length in Gothic, but I do not know his reasons for doing so.) The
one generative work that I have read in which vowel length is rejected is Wurzel 1975. His
reasons for doing so are essentially the same as those of the structuralists: “Wir können also
unterstellen, daß die Orthographie des Ulfilas die allgemeinen Prinzipien, die jeder überleg-
ten Neuschópfung von orthographischen Systemen zugrunde liegen, zur Richtschnur hat. . .
(1) Die Schreibung gibt phonetische Distinktionen dann und nur dann wieder, wenn sie kon-
trastieren. . . (II) Dieses Prinzip ist Grundlage für die Schaffung aller Buchstabenorthogra-
phien und bei neugeschaffenen orthographischen Systemen auch entsprechend befolgt. . .

284

(III) Die verwendeten Schriftzeichen werden dabei — falls sie einem bereits existierenden
Schriftsystem entstammen — soweit wie móglich mit dem phonetischen Wert benutzt, den
sie als Ánormalen' Wert in ihrem Herkunftssystem besitzen. . ." (266—7). Wurzel's biblio-
graphy suggests that he was not familiar with Penzl 1950; his “general principles,” however,
are essentially the same as Penzl’s.

The earliest generative work that I have read is Voyles 1968, and he assumes that vowel
length was relevant in Gothic. His argument is as follows: “In any event, Wulfila seems
generally to have disregarded length in the formulation of his orthography — except in the
single case of the representation of /i:/ by ei. This is perhaps ascribable to the prior ortho-
graphical traditions of fourth-century Greek and Latin on which Wulfila was in part depen-
dent, and which had no specific signs in general use to mark length or stress. Thus length is
not the only distinctive feature Wulfila chose to exclude from his orthography: he also
disregarded stress, which was doubtless phonemic in Gothic... At least one conclusion
may be drawn from this: Wulfila seems to have ignored the suprasegmentals in his other-
wise more-or-less phonemic analysis of Gothic phonology” (722). Except that I find it less
probable that an orthography will indicate stress than that it will indicate vowel length, I
find this argument very persuasive. At the same time, however, it is in a sense merely a
return to previous reasons for believing that vowel length was relevant in Gothic. In part-
icular, it is not based on any new insights derived from the methodology of generative
phonology. (Vowel length is also accepted in Voyles 1981.)

Very different from this is Vennemann 1971. Here the methodology of generative phono-
logy is used to provide evidence that there were two classes of vowels in Gothic, and that
they differed almost certainly as short vs. long. Though it is impossible to summarize in a
few paragraphs his long article (43 pages) and his very careful reasoning, I can at least
describe some of those parts of it which deal with vowel length.

It is generally accepted that Gothic shows a “lowering rule,” whereby certain vowels — but
not others — appear lowered before graphic h, h™, r. In order to give a generative accounting
for this, Vennemann suggests, let us assume a distinctive feature F (whatever it may turn
out to be) with plus and minus values; and let us assume that vowels which do not show
this lowering are *F, whereas vowels which do show this lowering are -F. Examples of these
two classes of vowels are as follows (insofar as possible, morphologically parallel forms are
given as illustrations; Roman numerals refer to the traditional classes of strong verbs):

Vowels with +F

Graphic Most environments Graphic Before h, AW, r

ei greipan (I, infin.) ei gateihan (I, infin.)
galeipan (I, infin.) leihWan (I, infin.)
skeinan (I, infin.) skeirs

u brukjan (infin.) u bruhta (pret.)
fuls skura

e qebun (V, pret. pl.) e frehun (V, pret. pl.)
gebun (V, pret. pl.) sehWun (V, pret. pl.)
nemun (IV, pret. pl.) berun (IV, pret. pl.)

o sokun (VI, pret. pl.) o slohun (VI, pret. pl.)

holop horos

285

Vowels with -F

Graphic Most environments Graphic Before h, hW, r

i undgripun (I, pret. pl.) ai gataihun (I, pret. pl.)
giban (V, inf.) saih Wan (V, inf.)
stiland (TV, pres. pl.) bairand (IV, pres. pl.)
gaswiltan (III, inf.) wairban (III, inf.)

u anabudun (II, pret. pl.) au tauhun (Il, pret. pl.)
andnumans (IV, ppl.) gabaurans (IV, ppl.)
gaswultun (III, pret. pl.) waurpun (II, pret. pl.)

(Vennemann gives far more examples than the above small selection, and the above ex-
amples are not necessarily the same as his — though they all illustrate the point he wishes
to make. To make sure that all of my examples are actually attested, I have checked them
with de Tollenaere & Jones 1976.)

The above examples show that graphic ei, e, o symbolized systematic phonemes with +F
(not lowered before A, AV, r), and that graphic i symbolized a systematic phoneme with
-F (lowered before h, hW, r). More interestingly, they show that graphic u symbolized two
different systematic phonemes: one with +F as in bruhta, skura (not lowered before h, r);
and one with -F as in tauhun, gabaurans, waurpun (lowered before h, r). Let us now
arbitrarily symbolize +F by writing a colon following the vowel symbol, and -F by not
writing a colon. This gives for Gothic the following two (partial) sets of systematic phon-
emes: /i: u: e: o:/ and /i u/.

A first question that we must now ask is: What was the phonetic nature of the distinctive
phonological features +F and -F? We cannot know for sure, of course. Diachronically,
however, /i: u: e: o:/ reflect earlier long vowels, and /i u/ reflect earlier short vowels. The
most likely answer is therefore: +F = +Long, and -F = -Long. (Vennemann also gives syn-
chronic phonological reasons for this assumption.)

A second question that we must ask is: Was this lowering rule a diachronic rule (no longer
operative in Wulfila’s Gothic), or was it a synchronic rule (still operative in Wulfila’s
Gothic)? Vennemann gives reasons for believing that it was “a very low-level process” (99)
and hence presumbably “synchronic” (130). Consequently, he does not list [e] and [9]
among the systematic phonemic vowels in native words of Gothic (126), because they can
be accounted for by the lowering rule; but he does list them among the systematic phonet-
ic vowels (130).

Vennemann summarized, as follows, what I take to be the basic theory underlying his
study: “Hypotheses concerning the segment inventory [of a language] have to be based not
on a study of the language as reflected in the spelling system, but on the grammar generat-
ing that language, especially its phonological component” (127). At the same time, he was
fully aware of the historical-comparative reason for assuming that vowel length was relevant
in Gothic: “It is not at all plausible that the Germanic language which was most amply
documented the earliest should at the same time be the most deviant one and should have
leveled a distinction which other Germanic languages have preserved to the present day”
(130).

The lively scholarly interest in the phonology of Gothic during the period 1948 to 1981 is
shown by the fact that I have been able to locate some forty studies devoted to this topic

286

during these years — and there may well have been more. Interpretations of vowel length
have changed as theoretical approaches have changed: first primary emphasis on the
historical-comparative evidence, then primary emphasis on the orthographical evidence,
then primary emphasis on the methodology of generative phonology. After such an ap-
parently exhaustive treatment, is it possible to add anything new? I believe that it is,
namely by using the approach of yet another theory.

The theory is that of “phonological space.” Though scholars have used this theory implicit-
ly for over a century and a half, remarkably little has been written about it explicitly (but
cf. Moulton 1962). Vowels, for example, occur in a phonological space that is formed by
such familiar “dimensions” as high vs. low, front vs. back, unrounded vs. rounded, short
vs. long, etc. The vowels in these interlocking dimensions form for any given language its
vowel “system”; and historical changes consist of the “movement” of vowels within the
system. As early as 1822 we find Jacob Grimm writing of “sounds” which “occupy a posi-
tion,” “move from a position,” “move down one step," and “fill a gap.” (For similar ideas
found also in the writings of Hermann Paul and Karl Luick, cf. Moulton 1962: 24—5.)
The theory of phonological space is also implicit in such common expressions as the
“lowering,” “raising,” “fronting,” “backing” of vowels. These terms would be meaningless
without the assumption of a phonological space in which such movements can take place.

Though these ideas were presumably well known, philologists did not in general use them
to group vowels into a “system.” At the beginning of this study I have given the vowels of
Gothic just as they are presented in Wright 1910 and Streitberg 1920: not as a “system”
but simply as a list — in an apparently arbitrary order. The notion of “system” was, how-
ever, obviously of fundamental importance in Trubetzkoy's famous 1929 article, "Zur all-
gemeinen Theorie der phonologischen Vokalsysteme”; and among the many languages used
as illustrations he included Gothic (126). In this first presentation of the vowels of Gothic
as a system, Trubetzkoy distinguished two classes of vowels: “maximalintensiv” (given
below to the left) and “minimalintensiv” (given below to the right). He displayed the
whole system graphicaliy as follows:

ä
au ai
ô ê :
- T u 1
ü i
(Trubetzkoy interpreted /au ai/ as falling diphthongs. Note that graphic iu does not appear

in this system.)

Then came the American structuralists. Since the notion that phonemes form "systems"
(or “structures”) was fundamental to their theory, they soon began to present the vowels
of Gothic as forming a system; an example is the system given in Marchand 1958: 508
(noted above), where the distinctive features that give the dimensions of the system are
fully specified. Then came the generative phonologists, who presented the vowels of Gothic
as a system of a very different sort: not as a set of elements occupying positions within
phonological space, but as a set of elements composed of distinctive phonological features.
Cf. Voyles 1968: 724, where the vowels (underneath them the distinctive features of which
they are composed) are listed linearly in alphabetical order — first short, then long. The
reason for this linear order is clear. The methodology of generative phonology is not

287

designed to describe sounds as elements of a system within phonological space, and in-
dividual sounds as elements which (to quote Jacob Grimm) “occupy a position” within
phonological space. In addition, the methodology of generative phonology was not designed
to describe sound change as the “movement” of elements within phonological space, where
the elements may (to quote Grimm again) “move from a position,” “move down one
step,” or “fill a gap.” (Without the theory of phonological space there can of course be no
such things as movements, positions, steps, or gaps.) Instead, sound change consists of
changes in distinctive features. This means that the methodology does not describe change
as such (“change-in-progress”) but only the results of change, as in the formula A > B/X
(feature A changes to feature B in the environment X). What this formula tells us is: before
the change, a given segment contained feature A; after the change, it contained feature B;
but the change as such is not described.

Vennemann (1971) seems to be an exception among generative phonologists in that
he accepts the notion of “system.” On p. 127 he gives the “‘vowel inventory of
Gothic” in the usual generative fashion: as a list of segments in a more or less arbitrary
linear order. However, on pp. 126 and 130 he displays first the two sets (short vs.
long) of systematic phonemic vowels and then the two sets of systematic phonetic
vowels as “systems” in a two-dimensional space composed (though not explicitly so
identified) of the dimensions high — low and front — back, in the usual configura-
tion: high at the top, low at the bottom; front to the left, back to the right. There is
only one slight violation of phonological space: {e 9] and [e: o:] are written as if
they were as low as [a] and [a:]. I welcome this implicit use, by a generativist, of the
theory of phonological space: it will simplify the remarks that I shall make shortly.

There are three phenomena in Gothic to which the theory of phonological space can be
applied so as to show that Gothic clearly had two sets of vowels — most likely distinguished
as short vs. long. All three phenomena have long been recognized. They are as follows:

(1) The lowering of earlier *i and *u before graphic h, hW, r, as described above. (Note
again that the very term “lowering” implies the theory of phonological space.)

(2) Graphic alternation between u and au, especially in u-stem nouns. This has been most
extensively investigated by Marchand (1956). He notes: 27 examples of au for expected u,
and 32 examples of u for expected au. (This is a gross simplification of his far more sophis-
ticated presentation.)

(3) Graphic confusion of ei, i, e. (Again I am simplifying Marchand’s very sophisticated
presentation.) We find:

(a) 65 examples of ei for expected e.
(b) 38 examples of e for expected ei.
(c) 20 examples of i for expected e.

(d) 16 examples of e for expected i.

(e) 5 examples of i for expected ei.
(f) 4 examples of ef for expected i.
(g) 10 examples of o for expected u.
(h) 4 examples of u for expected o.

Let us now consider how these three phenomena might be described in terms of the theory
of phonological space — where sounds are elements within a “system,” “occupy a posi-
tion” within the system, but may also “move” from one position to another within the
system. In each case I shall consider two alternative proposals: (A) that vowel length was

288

not relevant in Gothic; (B) that vowel length was relevant in Gothic. For (A) I shall use the
system explicitly described in Marchand 1958. For (B) I shall use the systems implicitly
presented in Vennemann 1971 (but I shall indicate /a/ and /a:/ as lower than /e »/ and
le: xN.

Let me repeat first the different correspondences between graphemes and phonemes as-
sumed in these two proposals:

(A) Length not distinctive: ei i e ai a auo u
li 1e e a 9 ouf

(B) Short: i ai a au u Long: ei e ai a auo u
lite ao uf li: eis: a:9: oru:

In proposal (B) I shall suggest that the long vowels were also tense, and hence farther from
the center of the spatial diagram; and that the short vowels were also lax, and hence closer
to the center of the spatial diagram. This is by no means a necessary assumption. The dia-
gram must merely show that, in addition to the dimensions high — low and front — back,
there was also the dimension short — long. I shall show this by placing the system of short
vowels inside the system of long vowels, and by drawing lines around it. I suggest that the
long vowels were tense and the short vowels lax merely because this is what we typically
find in systems with both long and short vowels. Vennemann (1971: 129) also considers it
“very likely" that the long vowels were (redundantly) tense, the short vowels (redundant-
ly) lax.

We can now consider how the three phenomena just mentioned might be described in the
phonological space of the single system of proposal (A) and in the phonological space of
the two systems (long and short) of proposal (B).

(1) Lowering of *i, * before graphic h, hW, r.
Proposal (A) Proposal (B)

In terms of the theory of phonological space, the lowering of *i, *u under proposal (A) is
highly implausible. First, the lowering of *i. On the one hand, we might assume (straight
arrows) that /1/ was lowered to and merged with /e/; we would then have to assume that all
words with /e/ somehow became disentangled so that only those from earlier /1/ were
lowered further to /e/, whereas those from earlier /e/ remained unlowered. Such a “disen-
tanglement” is quite implausible. Alternatively, we might assume (curved arrow) that, in
being lowered to /s/, earlier /1/ somehow bypassed intervening /e/. This, however, con-
tradicts the whole notion of “movement” within phonological space. The lowering of /u/

289

under proposal (A) is even more implausible. First, we would have to assume that the /u/
of some words was lowered, whereas the /u/ of other words was not. As for the words in
which /u/ was lowered, we have the same difficulties as above: either (straight arrows) a
lowering to and merging with /o/, and then somehow a disentanglement so that only the
[o] of words with earlier /u/ was lowered further to /»/, whereas the /o/ of all other words
remained unlowered. Or, alternatively (curved arrow), the /u/ of those words which show
lowering somehow bypassed the intervening /o/. This again contradicts the whole notion of
*movement" within phonological space.

In contrast, the lowering of "i, *u under proposal (B) presents no difficulties whatever.
Within the phonological space of the short vowels, /1 u/ simply “moved down one step,”
giving /e 2/. No more needs to be said.

(2) Graphic alternation between u and au.

Proposal (A) Proposal (B)

i i: u:
u

€ o e o:
At

€ I €: 3:

a
ai

The graphic alternation between u and au in terms of proposal (A) is, if anything, even
more implausible than the changes in (1). To explain au /5/ for expected u /u/ (downward
arrows), we would have to assume lowering to and merger with o /o/, then disentangle-
ment, then further lowering to au /5/. To explain u /u/ for expected au /5/ (upward arrows),
we would have to assume raising to and merger with o /o/, then disentanglement, then
further raising to u /u/. Alternatively, we would have to assume (curved double-headed
arrow) bypassing in both directions.

In contrast, the graphic alternation between u and au under proposal (B) is quite straight-
forward: it involves merely an alternation in height between two vowels that are neighbors
within short vowel space.

(3) Graphic confusion of ei, i, e, and u, o.
Proposal (A) Proposal (B)

ao :
O<———#

290

In terms of proposal (A), the graphic confusion of u and o is entirely compatible with the
theory of phonological space: it reflects merely a confusion between two vowels that were
neighbors within phonological space. (This is true, of course, only under the assumption
that graphic u reflects a single phoneme.) The graphic confusion of i with ei and e is also
entirely compatible with the theory of phonological space: /1/ moved sometimes one step
up to /i/, and sometimes one step down to /e/ (the upward and downward arrows). The
graphic confusion between ei /i/ and e /e/ is far more problematical. One might imagine
two “double confusions”: /i/ first moved one step down to and merged with /1/, and then,
after disentanglement, it moved a second step down to /e/ (the two downward arrows); and
/e/ first moved one step up to and merged with /ı/, and then, after disentanglement, it
moved a second step up to /i/ (the two upward arrows). Yet this seems hardly plausible.
There is, of course, the alternative of “bypassing” (curved double-headed arrow), though
this is incompatible with the whole notion of phonological space.

In contrast, the graphic confusion of ei, i, e and u, o under proposal (B) is quite unprob-
lematical (except that we do not understand why this confusion should have come about in
the first place). We find in every case a confusion between vowels that were neighbors
within phonological space. The confusion may involve two long vowels: /i:/ and /e:/, /u:/
and /o:/. Or it may involve a long vowel and a short vowel: /i:/ and /1/, /e:/ and /1/; /u:/
and /v/, /o:/ and /v/. To show that there is nothing unusual about a confusion (or fluctua-
tion) between a long vowel and a short vowel, we need to search no farther than English:
roof, hoof, etc. either with long /u:/ or with short /v/.

In view of the evidence from generative phonology (Vennemann 1971) and from the above
discussion of phonological space, it seems clear that (disregarding graphic iu) Gothic must
have had two classes of vowels, and that they were most likely distinguished as short vs.
long. We now face a question which plagues us all when we try to describe Gothic. Just
what do we mean by “Gothic”? Do we mean (1) pre-Wulfilian Gothic (at whatever period)?
Or (2) Wulfilian Gothic (mid 4th century)? Or (3) the Gothic of the scribes who wrote the
manuscripts (probably early 5th to late 6th century; cf. Marchand 1958: 497)?

Let us summarize what we know about vowel length at each of these three stages. The rule
lowering *i, *u before graphic h, hW, r was operative at stage (1), pre-Wulfilian Gothic
(and Vennemann suggests that it may have been operative down into Wulfilian Gothic).
It requires the assumption of vowel length. The graphic alternation between u and au may
have occurred during Wulfilian times, but it is more likely that it reflects stage (3), the lan-
guage of the scribes (cf. Marchand 1956: 148). It requires the assumption of vowel length.
The graphic confusion of ei, i, e and u, o almost certainly belongs to stage (3), the language
of the scribes (cf. Marchand 1956: 146 —7). It requires the assumption of vowel length.

This is probably the nearest we shall ever come to solving the riddle of vowel length in
Wulfilian Gothic. Two hypotheses are possible:

(a) Vowel length was relevant (1) in pre-Wulfilian Gothic, (2) in Wulfilian Gothic, and (3)
in post-Wulfilian Gothic. That is to say, proposal (B) was valid throughout the entire period.

(b) Vowel length was relevant (1) in pre-Wulfilian Gothic. (2) As the orthography shows,
it was not relevant in Wulfilian Gothic (this is proposal A). (3) Vowel length was relevant
in the Gothic of the scribes, though this was clearly a somewhat different kind of Gothic.

So what have we “proved”? Essentially nothing. But let the reader decide.

291
References

It is customary to list references alphabetically by author. Because so much of this study deals with the
history of views on vowel length in Gothic, it is more appropriate to list references chronologically by
date of publication.

1910. Wright, Joseph. Grammar of the Gothic language. Oxford: Clarendon Press.

1920. Streitberg, Wilhelm. Gotisches Elementarbuch. Sth & 6th edition. Heidelberg: Winter.

1929. Trubetzkoy, N.S. Zur allgemeinen Theorie der phonologischen Vokalsysteme. Travaux du Cercle
Linguistique de Prague 1: 39—67. Cited as reprinted in Josef Vachek, ed., A Prague School reader in
linguistics (Bloomington: Indiana University Press, 1964), 108—42.

1948. Moulton, William G. The phonemes of Gothic. Language 24: 76—86.

1949. Bennett, William H. The monophthongization of Gothic di du. Language 25: 15-21.

1950. Penzi, Herbert. Orthography and phonemes in Wulfila's Gothic. Journal of English and Germanic
Philology 49: 217—30.

1954. Penzl, Herbert. Review of Wolfgang Krause, Handbuch des Gotischen (1953). Language 30: 409—
14.

1955. Marchand, James W. Vowel length in Gothic. General Linguistics 1: 79—88.

1955. Marchand, James W. The sounds and phonemes of Wulfila’s Gothic. University of Illinois disserta-
tion. (Published 1973, see below.)

1956. Marchand, James W. Dialect characteristics in our Gothic Mss. Orbis 5: 141--51.

1958. Hamp, Eric P. Gothic ai and au again. Language 34: 359-63.

1958. Jones, Oscar F. Gothic iu. Language 34: 353-8.

1958. Marchand, James W. The Gothic language. Orbis 7: 492-515.

1962. Moulton, William G. Dialect geography and the concept of phonological space. Word 18: 23--32.

1964. Bennett, William H. Gothic spellings and phonemes: some current interpretations. Taylor Starck
Festschrift (London, The Hague, Paris: Mouton), 19-26.

1964. Buckalew, Ronald Eugene. A generative grammar of Gothic morphology. University of Illinois
dissertation. [Not seen.]

1968. Voyles, Joseph B. Gothic and Germanic. Language 44: 720—46.

1971. Beade, Pedro. Gothic phonology: a generative approach. Cornell University dissertation. [Not
seen.]

1971. Vennemann, Theo. The phonology of Gothic vowels. Language 47: 90—132.

1973. Beck, Richard. Length and monophthongization in Gothic. Indogermanische Forschungen 78:
113—490.

1973. Marchand, James W. The sounds and phonemes of Wulfila's Gothic. The Hague, Paris: Mouton.
(Janua linguarum, Series practica, 25.) (See above, 1955, Marchand.)

1975. Wurzel, W.U. Der gotische Vokalismus. Acta Linguistica Academiae Scientiarum Hungaricae 25:
263-338.

1976. de Tollenaere, Felicien, & Randall L. Jones. Word-indices and word-lists to the Gothic Bible and
minor fragments. Leiden: Brill.

1981. Voyles, Joseph B. Gothic, Germanic, and Northwest Germanic. Wiesbaden: Franz Steiner. (Zeit-
schrift für Dialektologie und Linguistik, Beihefte, Heft 39.).

Ein Halbvers und einige Epitheta aus vorhomerischer Dichtung*

Hugo Miihlestein (Université de Neuchatel)

Summary

This paper offers new observations on the Ethiopid as a source of the Iliad and on
epithets borrowed by Homer from his sources. The known facts: In the Ethiopid
Achilles’ friend Antilochus saved his father from Memnon’s attack and was killed by
Memnon and then revenged by Achilles. Surpassing the author of the Ethiopid,
Homer invented a new friend for Achilles, Patroclus, and a new adversary, Hector.
As Achilles, insulted by Agamemnon, does not fight and thus leaves his friend un-
protected, Patroclus is killed by Hector and revenged by Achilles.

Although the parallelism was noticed long ago, the priority of the Ethiopid remained
controversial until the origin of the personal names Patro-kl-os and Men-oit-ios was
clarified, thus proving it. In these names Homer evoked his pattern, namely the
heroic deed of Antilochus answering his father’s cry for help and standing firm, when
facing certain death.

With a view to imitating in his epic the great scenes of the death of Achilles and of
the following events, Homer transferred these motives to Patroclus as well. For his
death he curiously enough combined both his models, i.e. the death of Antilochus
and that of Achilles by Apollo and Paris. Therefore he created, debasing Hector’s
heroic role, two adversaries of Patroclus preceding Hector, Apollo and Eu-phorb-os,
the latter modelled on Paris, hence his shepherd’s-name.

Furthermore an epithet of Patrochus, innor&Xevdos ‘finding a way for the horses’,
also quotes the exploit of Antilochus, who deblocked his father’s chariot. Attributed
to Patroclus the word has no sense and was accordingly misunderstood by all the
interpreters both antique and modern, except by a single scholiast. Here is the crucial
point: That scholiast cannot have been able to understand the epithet from the Iliad
either. His knowledge must have derived from comments on the Ethiopid, and thus
the very word must have occurred in that poem.

Now we can conclude of its context there. On metrical grounds the epithet stood in
the vocative case, in a hemistich *’AvriAox’ immoke&Xevde. Such an address can only
have been the beginning of a challenge from Memnon to Antilochus immediately
before his last fight.

That Homer transferred an epithet from a model to its copy, without regard to its
meaning, furnishes a heuristic principle: Others of Homer’s epithets may now be in
the same class. Indeed xa\koKopvorns, kopvdaiodoc, and üàv6 poqóvoc can be proved
to be transferred from Memnon to Hektor, d0Adecoa and Sewn from Circe to Calyp-
so, immevs from Antilochus to Patroclus again. And Homer was not the only one to
act like this. The practice was usual among epic poets: In the pre-Odyssean Argonau-
tica for instance the Queen of the Phaeacians, Arete, was called AeuKwWevos after
Hera, her model.

* Diese Arbeit ist mit Zusätzen, die hier nicht mehr haben aufgenommen werden können, erstmals im
MUSEUM HELVETICUM 43, 1986:4 erschienen.

294

1. Unter den verschollenen Epen vom troianischen Sagenkreis nimmt die Aithiopis eine
Sonderstellung ein, weil sie, älter als Homers Epos vom Zorn des Achilleus, ihn, Homer, zur
überbietenden Nachbildung angeregt hat. Ein Kernstück der Aithiopis erzählte den Opfer-
tod des Antilochos, der seinen Vater Nestor vor dem heranstürmenden Aithiopenfürsten
Memnon gerettet hat, aber dabei selber vom weit überlegenen Memnon erschlagen wurde.
Ich wage nun zu vermuten, daß Memnon den Nestoriden vor dessen letztem Kampf ange-
sprochen und seine Rede mit den Worten *’AvriAox inmoke&Xevde eingeleitet hat. Hier ver-
suche ich, das glaubhaft zu machen, und werde anschließend die Herkunft einiger Epitheta
von Personen Homers aus den Vorlagen des Dichters nachweisen.

2. Zuvor muß an einiges bereits Gesicherte über Homers Inspirationsquelle für seine Menis-
dichtung erinnert werden. Die Erkenntnis, daß er durch die Aithiopis zu einem eigenen
Epos von Achilleus herausgefordert worden ist, hat sich gegen die seit Welcker! herrschen-
de Meinung, alle kyklischen Epen seien nachhomerisch, erst in unserem Jahrhundert durch-
gesetzt. Nach wenigen Vorläufern brachte vor allem das Buch von H. Pestalozzi” den
Durchbruch. Aber der schlüssige Beweis der Abhängigkeit Homers von der Aithiopis gelang
erst auf dem Umweg über die Namen der von Homer erfundenen Helden Patroklos und
Menoitios, die zusammen mit dem Wort inmok eXevdos auf die hohe Stunde des Antilochos
in der Aithiopis Bezug nehmen’.

3. Man weiß: Achilleus hat zwei Freunde, Patroklos und Antilochos, diesen schon in der
Aithiopis, jenen erst bei Homer, der ihn für sein neues Achilleusbild hinzuerfunden hat.
Beide Freunde fallen, beide werden von Achill gerächt. Aber in der Aithiopis fiel der
Freund in Selbstaufopferung für seinen Vater, ohne alle Mitschuld Achills. Dessen Helden-
tum war dort, daß er zur Rache an Memnon auszog, obwohl er wußte, daß er selber gleich
nach Memnons Tod würde sterben müssen; das hatte ihm seine Mutter Thetis auf Grund
von Vorzeichen geweissagt und damit erwirkt, daß er sich des Kampfes gegen Memnon bis
dahin enthalten hatte. Von solcher Art also war der Achilleus der Aithiopis, die von den
Alten einem Arktinos von Milet zugeschrieben wurde.

4. Homer dagegen, mit seinem Vorgänger ringend, ihn übertreffend, konzipiert einen an-
dern Achilleus, der nicht wegen Vorzeichen und Warnungen vom Schlachtfeld fernbleibt,
sondern aus Groll gegen Agamemnon wegen beleidigter Ehre, und der deshalb selber, in
seinem unversöhnlichen Starrsinn, durch seine Kampfenthaltung am Tod seines Freundes
mitschuldig wird. Für einen solchen Achilleus ersann Homer den Streit mit Agamemnon,
die erfolglose Bittgesandtschaft und eben, die Rollen der Vorlage frei nachbildend, den
neuen Freund und den neuen Gegner, also an Stelle von Antilochos und Memnon jetzt
Patroklos und Hektor. Dabei dichtete Homer so sehr in Auseinandersetzung mit dem Epos
seines Vorgüngers — oder war jener Dichter gar noch sein Zeitgenosse, sein Rivale im mu-
sischen Agon? —, daß er Achills neuem Freund Namen, Vaternamen und Beiwort im Ge-
danken an den Opfertod des früheren Freundes gab.

! Der epische Cyklus IL (1849) 261 T.
? Die Achilleis als Quelle der Ilias (1945).
Studi Micenei ed Egeo-Anatolici 9 (1969) 86 -93 (mit Literatur S. 91 Anm. 24).

295

5. Wir kennen nämlich, vor allem dank Pindars rihmender Erwähnung? , den Hergang jener
Heldentat des Antilochos in der Aithiopis ziemlich gut: Paris hatte ein Pferd von Nestors
Gespann angeschossen, Nestor war blockiert und hatte, vom heranfahrenden Memnon be-
droht, um Hilfe gerufen. Antilochos hörte auf den Ruf seines Vaters — etwa matpos éx Ave —,
eilte, aus Pietät allzu kühn, während die andern Achäer flohen, herbei und bereitete dem
Wagen einen Ausweg — innowı keXevdov —, indem er die Riemen des verletzten Pferdes
durchhieb und es dadurch vom Wagen abtrennte, so daß Nestor wegfahren konnte. Dabei
wagte es der Sohn, um den Rückzug des Vaters zu decken, vor dem gewaltigen Memnon
nicht zu fliehen, sondern sich ihm in den Weg zu stellen und stehen zu bleiben vor dem
sicheren Untergang — uevew otrov —, und fiel. Deshalb also gab Homer Achills neuem
Freund den Namen Ilárpo-kXoc? , das Epitheton immo-k&Aevdos und einen Vater Mev-oir-
wos.

6. Es ist denn auch ganz unmöglich, diese Namen und dieses Epitheton etwa auf Patroklos
selber sinnvoll zu beziehen. Denn es gibt in Homers Patroklie keine Stelle, wo der Held auf
seinen Vater hörte®, keine, wo er auf den Tod gefaßt einem mächtigeren Gegner widerstün-
de, keine, wo er gar für irgendwelchen behinderten Kriegswagen einen Ausweg fände. Die
Gegenprobe bestätigt es also: Die drei Wörter weisen auf die Vorlage zurück, auf Antilo-
chos und seine große Tat. Nach diesem Befund ist die Umkehrung, die Annahme einer
Priorität des Patroklos vor Antilochos, also der Menis Homers vor der Aithiopis im Sinne
Welckers und führender Gelehrter nach ihm, nicht mehr denkbar. Aber erstaunlich bleibt
dieses Verfahren Homers, hier wie anderswo, seinen Personen Namen und Beiwörter nach
ihren Vorbildern zu geben.

7. So hat denn Homer, was er von Achill, Patroklos und Hektor erzählt, zwar zeitlich nach
dem erfunden, was die Aithiopis von Achill, Antilochos und Memnon erzählte, aber seinen
neuen Achilleusstoff im Ablauf des ganzen Achilleusgeschehens notwendig den Inhalten
der Aithiopis vorangestellt; denn Achills Ende unmittelbar nach seiner Rache an Memnon
war dank dem Erfolg der Aithiopis in die Sage eingegangen und zur gültigen “Wahrheit” ge-
worden, wie eben ‘Sage’ entsteht: aus großer Dichtung, die im Bewußtsein der Hörerge-
meinde sich einprägt, für wahr gehalten wird und so schließlich unaufhebbare Gültigkeit
erlangt. Homer mußte also die Episoden für sein neues Achilleusbild sagenchronologisch
gleichsam rückwärts erfinden und konnte in seiner Menis mit Rücksicht auf die Aithiopis
besonders nicht nochmals — und anders! — den Tod des gleichen Achilleus erzählen.

8. Die Einfügung des Neuen ins Alte ist übrigens nicht ganz schlackenlos gelungen. Folge-

richtig hätte Homer seinen Achilleus, der für den Kampf mit Memnon überleben mußte, zur
^ Pythia 6, 28ff.

2 Daneben die vollen Formen llarpokAéeoc, -kAéea, -kAeec, also ursprünglich etwa ‘Des Vaters Ruhm
habend’, ‘Berühmt wie sein Vater’. Aber Homer hat den Namen auf das mit kAéoc verwandte xAveiv
umbezogen, ähnlich wie den Namen der Fopv-kAeta, die nicht Weithin berühmt’ heißt, sondern ‘Der
man weithin gehorcht', vgl. u 147ff). Diese Namen hat der Dichter eben nicht selber erst gebildet,
sondern aus dem reichen schon bestehenden Namenschatz zu seinen poetischen Bedürfnissen passend
ausgewählt und frei gedeutet.

Denn Nestors Worte A 785 790, mit denen er Patroklos an die Mahnungen seines Vaters erinnert,
auf die der Sohn zwar nicht hört, aber hören sollte, gehören nicht Homer, sondern einem Bearbeiter
(vgl. Von der Mühll, Kritisches Hypomnema zur Ilias 200), der sich jedenfalls gerade vom Namen für
das Motiv hat inspirieren lassen.

296

Rache an Hektor nicht in der Meinung ausziehen lassen dürfen, er werde diese Rache sofort
mit dem eigenen Tod bezahlen müssen; denn das geschieht ja nun gar nicht sofort. Und
doch läßt Homer die Mutter es ihrem Sohn so voraussagen, 2 96: abrika yap Toı Ereıra
ue “Exropa nôruos Erowos (vgl. schon 89ff und dann 115f). Aber Homer hat die leichte
Inkonsequenz in Kauf genommen, um das auch für sein neues Achilleusbild unentbehrliche
Motiv der Vorlage, Achills Wissen um den Tod als Preis der Rache (s.o. Abschnitt 3), nicht
zu verlieren.

9. Nun haben aber gerade auch die Ereignisse der Aithiopis, die auf Achills Rache an
Memnon folgten, Homer zur dichterischen Überbietung herausgefordert. Das war Achills
Sturm auf die Stadt, sein Tod durch Apollon und Paris im Skäischen Tor, dann Kampf um
die Leiche, Bergung, Aufbahrung, Totenklage durch Thetis mit ihrem Gefolge, Bestattung
und Leichenspiele. Um auch diese großen Motive in seine Menisdichtung hereinzuholen, hat
Homer mit einem genialen Kunstgriff sie alle vom Achilleus der Vorlage noch auf den von
ihm erfundenen Patroklos übertragen. Dieser spiegelt also den Antilochos und den Achil-
leus der Aithiopis in einer Person.

10. Aber diese Verschmelzung ist, wenigstens an der Stelle, wo Patroklos fällt, nicht nur
zum Vorteil der Dichtung ausgefallen. Homer hat nämlich seltsamerweise für den Tod sei-
nes Patroklos nicht etwa den Tod nur des einen der beiden Vorbilder nachgezeichnet, son-
dern die Tode des Achilleus und des Antilochos aus der Vorlage gleichsam aneinanderge-
hängt. Das ergibt eine unbefriedigende Szene und verkleinert namentlich Hektors Helden-
rolle. Denn bevor dieser den Patroklos erlegt, II 818ff, gedichtet auf dem Hintergrund von
Memnons Sieg über Antilochos, hat ihn schon Apollon gelähmt und entwaffnet und ein erst
hier eingeführter Euphorbos in den Rücken geschossen, II 785ff, gedichtet auf dem Hinter-
grund vom Tod Achilis durch Apollon und Paris, so daß Hektor nun nur noch einem Wehr-
losen den Rest gibt, wahrlich keine Ruhmestat. Es ist erstaunlich, daß Homer nicht schon
selber an der Unvereinbarkeit der beiden Tode Anstoß genommen und den Ruhm, Patroklos
besiegt zu haben, nicht ungeteilt seinem Hektor gelassen hat, den er doch sonst besonders
liebevoll zeichnet. Erfand der Dichter unter einem inneren Zwang, von seiner Vorlage
nichts fallen zu lassen, alle ihre Motive einzufangen und konkurrierend nachzubilden?

11. Wie dem auch sei, unvollkommen ist auch die Erfindung des bis dahin unbekannten
Euphorbos für seine kurze Rolle. Ja, die Vorgeschichte des Mannes, I1 808—811, wird erst
verständlich, wenn man weiß, wem er Zug um Zug nachgebildet ist. Und auch hier wieder
hat der Name einen ersten Hinweis auf das Vorbild gegeben: Auf den Hirtennamen Eö-
wopBoc, ‘Mit guter Weide', ist der Dichter beim Gedanken an die berühmte und folgen-
schwere Hirtenzeit des troianischen Königssohnes Paris im Idagebirge gekommen’. Übri-
gens zeigt gerade der Name Euphorbos, der die Aussetzung des Paris und damit das Parisur-
teil voraussetzt, zusammen mit weiteren, hier nicht zu erörternden Indizien, daß Homer
auch andere Inhalte des epischen Kyklos kannte. Aber sein Gegenstand war der Held
Achilleus, wie er ihm in der Aithiopis entgegentrat und ihn zur Zeichnung eines anderen,
gewaltigeren Charakters herausforderte.

7 Ausführlich über Euphorbos in Studi Micenei 15 (1972) 79ff. — W. Schadewaldt, Von Homers Welt
und Werk (19522) 173, hat das doppelte Vorbild für Patroklos in Antilochos und Achilleus erkannt
und in einem Schema veranschaulicht, aber noch ohne die Erfindung des Euphorbos zu verstehen.

297

12. Denn die Namen beweisen jetzt eben nicht nur, was die Motivanalyse zuvor schon
wahrscheinlich gemacht hat, die Priorität der Aithiopis vor Homers Menis, sondern auch,
daß Homers Achilleusdichtung, dieses Kernstück noch unserer Ilias, überhaupt erst im Rin-
gen mit jener Vorlage konzipiert und durchgebildet worden ist. Wie frei Homers Genie da-
bei gegenüber seiner Inspirationsquelle blieb, erkennen wir besonders deutlich an dem, wo-
zu ihn die Penthesileia-Episode der Aithiopis? angeregt hat. An Stelle der von fraulicher
Wirklichkeit so weit entfernten Kriegerin und Anführerin zeichnete er, vermenschlichend
nach seiner Art, ein diametral entgegengesetztes, dafür in seiner und in jeder Welt sehr rea-
les Frauenbild: des Kriegers Gattin — Hektors Gattin. Von dem Vorbild — oder “Antivor-
bild" — steckt eine Spur noch in Andromaches taktischem Ratschlag an den Gatten, Z
433ff, den er sanft zurückweist, 490ff: “Walte du nur im Haus, der Krieg bleibt Sache der
Männer, 7ÓAeuoc 6 ürópeoot ueinceı”, im Sinne des Dichters, der die Amazone nicht
mochte. Der Beweis aber von des Dichters Anregung durch die Penthesileia der Vorlage
steckt, einmal mehr, im Namen der ‘Avépo-uaxn. Dieser ist nicht nur nach seinem Sinn,
“Die gegen Männer Kämpfende’, ein Amazonenname, sondern auch in der Literatur und auf
Vasenbildern, und dort der bei weitem häufigste.

13. Warum Homer oft die Namen seiner Personen nicht auf diese selber, sondern auf deren
Vorbilder hat hinweisen lassen, ist wohl schwer zu sagen. Immerhin stellen wir diese Art
der Namengebung nicht nur bei ihm fest, sondern auch bei andern in unseren Homertext
eingearbeiteten Epikern, z.B. beim Dichter der Ithakesierversammlung im ß, der erst um
560 in Athen an der Odyssee weitergedichtet und in den Namen der dem Odysseus treuen
Greise unter den Rednern auf den alten Solon, in denen der frechen Freier auf den Usur-
pator Peisistratos angespielt hat”. Das spricht dafür, daß diese seltsame Benennungsweise
zur überlieferten epischen Techne gehörte. Ein Wille des Dichters, durch diskrete Hinweise
den kundigeren seiner Hörer die Quellen und treibenden Kräfte seines Schaffens zu verra-
ten oder seinen Vorgängern ein verschlüsseltes Denkmal zu setzen, ist kaum anzunehmen,
zumal die antiken (und die modernen) Gelehrten solche Zeichen in den Namen gar nicht
bemerkt zu haben scheinen.

14. Was auch immer einen Dichter zum Evozieren von Vorbildern in den Namen bewegen
konnte, daß Homer und seine Erweiterer so gedichtet haben, ergibt für uns eine Forschungs-
methode. Auf diesem Weg kann z.B. die vielerörterte Frage nach dem Verhältnis des
Achilleuszornes zum Meleagroszorn entschieden werden, von dem der Dichter in Umbie-
gung älterer Sage den Phoinix im Paradeigma für Achilleus erzählen läßt, I 524-599: Die
Rolle der KAeo-narpn, 556ff, geht mit der des Patro-klos parallel, und die Namen sind aus
den gleichen Elementen zusammengesetzt, also der eine aus dem andern durch Umstellung.
Aber welcher ist des andern Vorbild? Nachdem klar ist, daß Patroklos nach der Heldentat
des Antilochos benannt ist, kann nur Kleopatre nach Patroklos gebildet, der Meleagroszorn
in Analogie zum Achilleuszorn erfunden sein’®.

® Noch Pestalozzi S. 5 und Schadewaldt a.O. S. 158 glaubten, zu Unrecht, die Penthesileia-Episode sei

“offensichtlich jünger”, bzw. “sichtlich ein anderes Gedicht” als die Memnon-Episode. Die Behaup-
tung, eine Liebe Achills zu Penthesileia könne erst lange nach Homer erfunden worden sein, ist will-
kürlich (richtig W. Kullmann, Die Quellen der Ilias 46f).

Studi Micenei 17 (1976) 137—155.

10 Studi Micenei 9 (1969) 86f und zuletzt Ziva Antika 31 (1981) 87f.

298

15. Nach den Namen nun zu den Epitheta, und vorerst zu inmokeXevdos (s.o. Abschn.
1.2.4ff). Denn auch dieses Wort verstehen wir erst, nachdem die Namen erwiesen haben,
welche Szene der Aithiopis dem schaffenden Dichter vorschwebte, s.o. Abschn. 5. Es ist
formal ein Possessivkompositum, “für die Pferde einen Weg habend’, auf Antilochos bezo-
gen ‘dem Wagen einen Ausweg findend’, und hält so, in knappster Zusammenfassung, ein
sehr besonderes, beispielloses und kaum wiederholbares Ereignis fest, einen Auftritt, für
den allein es geschaffen worden ist, ohne Aussicht, je wieder in seiner richtigen Bedeutung
Verwendung zu finden. Aber schon in seinem einzigen erhaltenen antiken Vorkommen, bei
Homer in der Formel IlarpoxAees innoreXeude, II 126.584.839, verliert es, auf das Abbild
bezogen (s.o. Abschn. 6), allen Sinn. Welche Zumutung Homers an sein Publikum, das in
seinem Gesang von Patroklos keine Episode zu hören bekam, wo dieser einen blockierten
Wagen wieder in Fahrt gebracht hätte! Man konnte das Wort im Zusammenhang der Menis
überhaupt nicht verstehen.

16. So ist es — mit einer noch aufzusparenden Ausnahme — auch für die Erklärer geblie-
ben, antike und moderne. Ich zitiere Scholien, Übersetzungen, Lexika und Kommentare
nicht einzeln. Sie erklären immokeXevde als inmike; Taxurare; &p' innwv nowúueve TV
Kédevdov; napanınolav inmors Opunv Exwv; ‘qui combat sur un char; que transportent
les chevaux; travelling by means of horses; driver of horses; rider; Rossetummler; einen lan-
gen Weg fahrend’; auch, mit falscher Ableitung von KeAevw, ‘Pferden befehlend’. Auch wo
das Wort frühbyzantinisch noch einmal in einem Gedicht erscheint, Anth. Pal. 9, 210, 9,
steht es katachrestisch für ‘Reiter’. Bentley fand das Wort zwar mit dem Element K&Xevdos
sinnlos im Zusammenhang, aber in der überlieferten Form auch nicht auf das ihm sinnvoller
scheinende keXevw zu beziehen — beides mit Recht —und wollte deshalb in *inmoreAevora
korrigieren!!. Alle diese Deutungen tun der Wortbildung oder der Überlieferung oder dem
Sinn von kéAev90c Gewalt an.

17. Aber nun findet sich unter den antiken Erklárungen doch auch einmal, schol. I1 126,
die richtige: roi inno eürpemigcov 060v, ‘dem Gespann einen Weg bereitend’, und das
führt weiter. Hätte nämlich der Interpret, tüchtiger als alle seine Kollegen, das Wort im
Homer richtig als Possessivkompositum erkannt, nur nach seiner Form und trotz seiner
Sinnlosigkeit dort, so hátte er nicht ebrpemücov, sondern éxcv geschrieben. ‘Einen Weg be-
reitend' dagegen erklärt ausdrücklich das Verhalten des Antilochos, und dieses kann der In-
terpret bei der Einmaligkeit jener Tat!? letztlich nur aus der Aithiopis gekannt haben, sei es
direkt aus dem noch erhaltenen Epos, das Vergil noch gelesen zu haben scheint, sei es aus

Glossarien, in welche die Deutung des so ungewöhnlichen Wortes aus einem Kommentar
! E, Bechtel im Lexilogus zu Homer s.v. hat îimmokéAevdos als poetische Bildung erkannt und bekennt
die Unverständlichkeit des Ausdrucks. Auch D. Page, History and the Homeric Iliad (1959) 256 und
286 Anm. 90, hat die Unvereinbarkeit des Epithetons mit der Rolle des Patroklos notiert, aber nicht
mit dem Vorbild aus der Aithiopis und der es verratenden Aussage der Namen zusammengesehen, so
daß er, zu Unrecht, noch an einen vorhomerischen Patroklos denken konnte, der had much to do
with horses”; er glaubte, das Wort gehöre zu keAeUqw.

Denn das, was Apollon O 260f dem Hektor verspricht, inrowoı keAevdov racav Aciavéw, also mit
den Elementen des Wortes inmoreAevdos, und 355-366 ausführt, indem er den Graben des Achäerla-
gers zuschüttet und den Wall schleift, um den Wagen eine Bahn zu Öffnen, ist doch etwas sehr ande-
res, als was Antilochos am Wagen seines Vaters tat, um ihn zu deblockieren. Übrigens ist Apollons
Einsatz im O mit starken Gründen einem Bearbeiter zugeschrieben worden, vgl. Von der Mühll, a.O.
230.

299

zur Aithiopis eingegangen war. So oder so, das Wort hat in der Aithiopis gestanden, Homer
hat es also nicht etwa selber im Gedanken an Antilochos erst neu gebildet!*, sondern fertig
aus der Vorlage übernommen und ohne Rücksicht auf seinen Sinn auf Patroklos übertragen.

18. Dieser punktuelle Befund erlaubt Schlüsse sowohl auf die Aithiopis als auch auf Ho-
mers Art zu dichten. Man kann abschätzen, wo in der Aithiopis das Wort irrnokéAeudos ge-
standen haben muß. Die Möglichkeiten lassen sich nämlich einschränken, zunächst me-
trisch: Als Epitheton des Antilochos stand das Wort vor oder nach diesem Namen und mit
diesem zusammen in einem der fünf Casus, aber von den zehn Kombinationen ist im Hexa-
meter eine einzige leicht einsetzbar, der Vokativ in der Abfolge Name — Beiwort am Vers-
anfang. Damit ist der eingangs formulierte Halbvers, 'Avri\ox’ innokekevde, gegeben.
Auch ist sein Ort in der Erzählung nun eng eingegrenzt; denn so anzureden war der Held
erst nach seiner Rettertat und eigentlich nur noch durch seinen Besieger Memnon, kurz vor
oder kurz nach ihrem Kampf, vgl. II 830 bzw. 859. Im Unmut über den unerwarteten Ver-
lust eines scheinbar sicheren Sieges über den berühmten Nestor macht Memnon das, was er
soeben als unerhórt kühne Tat mitangesehen hat, zum eigentlichen Wesen des Jünglings, der
sich ihm unvermutet und ohne alle Aussicht, ihm gewachsen zu sein, in den Weg stellt, und
faßt dessen Tat, extrem verdichtend, in das einzigartig gebildete Wort — ‘“du Streitwagen-
deblockierer!” Ein so kunstvolles Epitheton hat nur ein hochbegabter Dichter prägen kón-
nen. Schon er hat offenbar ein bedeutendes Epos geschaffen. Sonst hátte kaum ein Homer
sich davon zur Überbietung reizen lassen.

19. Nun also zu Homer. Wir haben aus seiner seltsamen Art, mit dem Namen des Patroklos
auf sein Vorbild für diesen, auf den Antilochos der Aithiopis, hinzuweisen, verallgemei-
nernd eine Methode gewonnen, um von Namen homerischer Personen auf ihre Vorlage zu
schließen, z.B. von Euphorbos auf Paris, von Kleopatre auf Patroklos, von Andromache auf
Penthesileia, natürlich immer nur im Zusammenspiel mit andern Indizien. Wir werden nun
aus Homers noch seltsamerer Art, das Epitheton inmoxreXevdos von jenem Antilochos sach-
lich unzutreffend auf dessen Abbild Patroklos zu übertragen, wieder verallgemeinernd noch
eine andere Methode gewinnen, die dem gleichen Ziel dient. Denn jenes Epitheton ist nicht
ein vereinzelter Fall. Vorerst ein Beispiel aus der Odyssee:

20. Homer nennt die Nymphe Kalypso 'tückisch', 60A0eooa (1 245), was sie, die liebende,
verstehende, verzichtende, mit keiner Faser ihres Wesens ist. Das Epitheton hat Homer,
auch hier wieder trotz dessen unpassender Bedeutung, auf seine Neuschópfung übertragen
von der Gestalt, die ihm für die Erfindung der Nymphe eine Vorlage hergab, námlich von
der Kirke, die so bezeichnet wird (1 32) und so war, wenn sie die Ankömmlinge in Schwei-
ne verwandelte. Nicht viel besser paßt auf Kalypso in der nächsten Zeile (n 246) dewn
deoc, ‘schreckliche Göttin’, nach dem Formelvers für Kirke (x 130 = À 8 = u 150). Die an-
dern Indizien für dieses Verhältnis der beiden Frauengestalten, aus deren göttlicher Liebe
Odysseus weg- und zu Penelope zurückstrebte, hat U. v. Wilamowitz schon vor hundert Jah-

3 Ich korrigiere hiermit meine in Studi Micenei 9, 91 geäußerte frühere Ansicht.
14 Nicht wohl am Versende; denn die vor dem Vokativ dann anzunehmende Zäsur fiele in eine dafür
nicht beliebte Stelle und würde den Vers halbieren.

300

ren notiert, Homerische Unters. 115ff, nicht widerlegt durch H. Giintert, Xa/ypso (1919).
Wir erkennen in Kalypso gegeniiber Kirke wieder, wie in Andromache gegeniiber Penthesi-
leia, Homers Willen und Kunst, das Unwirkliche und den Zauber durch ein menschlich
Wahres und Wesentliches zu ersetzen und zu überbieten.

21. Wir haben an den Epitheta innoké\evdoc, 6oAdeooa, Gen die Entlehnung vorerst des-
halb beachtet, weil diese Wôrter im neuen Kontext nicht oder schlecht passen. Aber Ho-
mers eigenartige Praxis, Epitheta von Vorbildern auf seine Neuschôpfungen zu übertragen,
war natürlich nicht auf solche beschränkt, die dann nicht paßten. Im Gegenteil, die meisten
so entlehnten blieben gewiß auch im neuen Zusammenhang sinnvoll oder wenigstens nicht
sinnwidrig. Um auch solche festzustellen, achten wir also auf andere Indizien als nur auf
ihre Unangemessenheit. Es kann z.B., umgekehrt, ein Epitheton zwar im neuen Kontext
nicht sinnlos, aber auf das Vorbild ganz besonders zugeschnitten sein. Oder es ist bei
Homer ausschließlich (oder fast ausschließlich) einer einzigen Person beigegeben, obwohl
es zu andern Personen ebenso gut passen würde. Oder es ist als Beiwort auch des Vorbildes
bezeugt, direkt oder indirekt. Die Kumulierung mehrerer dieser Indizien bei einem einzel-
nen Epitheton führt dann zur Gewißheit, zumal wenn mehrere Epitheta eines Vorbilds auf
sein Abbild übertragen sind.

22. Man wird beachten, daß schon bei den drei hiervor besprochenen Wörtern je mehrere
dieser Indizien die Entlehnung verraten. Wir suchen jetzt solche entlehnten Epitheta unter
denen Hektors, den Homer in seiner Menis als Sieger über Achills Freund und Opfer von
Achills Rache aus der Rolle Memnons in der Aithiopis heraus erfunden hat, s.o. Abschn. 4.
Von den zahlreichen Beiwörtern Hektors bespreche ich hier drei, die sich mit Sicherheit
als Entlehnungen von der Vorlage erweisen.

23. xaAkokopvorns 'erzbehelmt, erzbewaffnet' ist achtmal Epitheton Hektors, einmal
(Z 199) Sarpedons. Auch der ist ein Abbild Memnons; denn Homer hat für beide konstitu-
tiven Elemente seines Patroklos je den gemäß der Aithiopis zugehörigen Gegner eingeführt,
für das antilochische den Hektor, der ihn besiegt wie Memnon den Antilochos, für das
achilleische den Sarpedon, den er besiegt wie Achill den Memnon. Die Beschränkung von
XxaAKoKopvornSs auf diese beiden Helden, obwohl auch die andern erzbehelmt sind, ist ein
erstes Indiz für die Übertragung des Wortes vom Memnon der Aithiopis auf Hektor und
Sarpedon. Hinzu kommt, zweitens, daß es bei Memnon besonders angezeigt scheinen muß-
te, in einem Epitheton seine Rüstung zu evozieren; denn er kam, wie noch der Auszug aus
der Aithiopis bei Proklos es festhält, ëxwv hyarorórevkrov navonNiav den Troianern zu
Hilfe. Die Übertragung wird zur Gewißheit, wenn, drittens, in Hesiods Theogonie 984,
letztlich aus der gleichen Quelle, Memnon selber das Beiwort erhält: Tawa 8’ "Has
Téke Méuvova XaAKOKOPVOTND.

24. Kopvdaloros ‘helmblitzend’!° ist achtunddreiBigmal Beiwort Hektors, ein einziges Mal
nicht, Y 38 von Ares in der erst nachhomerischen Theomachie. Bei Homer gehört es also
15 Aber v. Wilamowitz schloß daraus zu einfach (116), daß “es folglich eine Zeit gab, wo Odysseus zwar
bei Kirke war, aber nicht bei Kalypso.” Welcher Held bei Kirke geweilt hat in jener (korinthischen?)
Quelle, aus welcher ein späterer Bearbeiter (der letzte?) die Zauberin zum Glück noch in unsere
is Odyssee hereingeholt hat, bleibt ungewiß (lason?).
Nicht ‘helmschüttelnd’; s. die Diskussion bei Page, a.O. 249f und 288f Anm. 92 —95.

301

nur dem Hektor. Diese Ausschließlichkeit und die Bedeutung des Wortes, das wie xaAkoko-
pvorns zu den Hephaistoswaffen Memnons paßt, zeigen deutlich, daß auch dieses Epithe-
ton von Memnon auf Hektor übertragen ist, zumal es mit den andern Übertragungen von
Memnon auf Hektor und von Antilochos auf Patroklos parallel geht. Bei Memnon stand es
freilich noch nicht wie bei Hektor in einer Versschlußformel’”.

25. àvôpopôvoc ‘männermordend’ ist elfmal Beiwort Hektors, einmal des Ares, einmal des
Lykurgos, und dreimal bezeichnet es die Hände Achills. Die Bedeutung paßt zu jedem Krie-
ger. Umso auffälliger ist die Häufung des Epithetons bei Hektor unter Ausschluß so vieler
andern. Das würde kaum genügen, um das Wort für Hektors Vorbild zu beanspruchen, käme
nicht hinzu, daß Pindar in seinen Versen über des Antilochos Tat, Pythia 6,30, für Memnon
das mit avd popdvos synonyme Evapiußporos passend gefunden hat, evident als Paraphrase
von Formeln wie *Méuvoroc àv6 popóvoto aus der Aithiopis!? .

26. Zuriick zu Patroklos. Neben dreimaligem HarpdkAXeec immox éNevde (s.0. Abschn. 15)
steht konkurrierend in der Patroklie viermal HarpdkXees immed!”. Homer braucht imtevs
sonst nur im Plural, nie als Epitheton. Als solches stand es zweifellos in der Aithiopis bei
Antilochos, dem Vorbild, etwa in einer Versschlußformel *’AvriAox’ immed. Zu diesem
paßte es besonders gut, weil gerade Nestor und seine Söhne als Streitwagenkrieger und auch
sonst als Fahrer berühmt waren.

27. Denn so viel von pylischer Tradition ist noch faßbar. Daraus schöpft der letzte Bearbei-
ter Homers, in Athen im Dienst und zu Ehren der Peisistratiden, wenn er Nestor im A,
720—761, mit seinen Jugendtaten als Wagenkämpfer prahlen, wenn er ihn im v, 304—348,
seinem Antilochos Anweisungen für das Wagenrennen geben, wenn er ihn im A, 510ff, mit
seinem Wagen den verwundeten Machaon aus der Schlacht retten, auch wenn er, y 482ff,
den andern Nestorsohn Peisistratos sich im Fahren auszeichnen läßt. Aus der gleichen Tra-
dition schópfte schon lange zuvor in Ionien, wo man sich ebenfalls auf pylische Herkunft
berief”, der Aithiopisdichter, wenn er den Nestoriden Antilochos in der Streitwagenbe-
drängnis seines Vaters sein Heldentum beweisen ließ. Der Tepnvios imnora Neorwp ist eben
in mindestens zwei Malen in die troianische Epik aufgenommen worden, einmal vor und
einmal nach Homer, und was fiir den Vater immora, das war für den Sohn izmevc. Nicht zu-
fállig galt der Gott der Pferde, Poseidon, als Ahn der Neleiden, des pylischen Geschlechts,
aus dem auch die Peisistratiden sich herleiteten?! .

28. Die hier vor allem am Verhältnis von Homers Menis zur Aithiopis dargestellte seltsame
Art, mit Personennamen Inspirationsquellen zu evozieren und Epitheta von Vorbildern
17 Auf Memnon und seine Hephaistoswaffen geht vielleicht auch Hektors Beiwort yaißıpos zurück
(29mal). Der Glanz der von Hephaistos geschmiedeten Rüstung ist bei Homer ein unterscheidendes
Merkmal auch Achills, wenn Priamos diesen daran erkennt und der Dichter diesen Glanz mit dem des
Sirius vergleicht, X 24--32. Aber gaid quos steht vierzehnmal auch bei andern Helden.

Ähnlich hat Bakchylides ein überliefertes Beiwort durch ein Synonym ersetzt, wenn er die in ätoli-
scher Epik ursprüngliche Formel inmnAdra Oivevs (I 581, vgl. immnAdra Tvdevs A 387) für seine
Daktyloepitriten in nAa£ınnos Oivevs abwandelte, s. Studi Micenei 9,85; Ziva Antika 31 (1981) 89
Anm. 12.

T1 20. 744.812. 843.

Mimnermos 12 D.

Herodot 5, 65,3.

302

- mechanisch auf Neuschöpfungen zu übertragen, zeigt Homer nicht nur in der Ilias. In der
Odyssee, s. oben Abschn. 20, überträgt er Epitheta von Kirke auf Kalypso. Der Name des
HoAv-pnuos, Synonym der Hadesepiklese Klymenos, verrät, woran Homer bei der Erfin-
dung dieses Poseidonsohnes dachte”. Der ‘göttliche Sauhirt’ Eumaios hat seinen Namen
nach dem quasi-synonymen des wohlwollenden Unterweltsgottes Eubuleus, der nach der
Legende ebenfalls Schweinehirt und Helfer der Demeter war?” und dem Dichter den treuen
Diener des Odysseus eingegeben hat. Auf die gar originell wirkende Verbindung dios
bpopßos (auch Si” Eünare) kam Homer, weil er den Gott im Kult als Zeus Eubuleus kannte,
und zum Vater gab er dem Eumaios einen Krnows (o 414) nach dem ebenfalls unterweltli-
chen Zeus Ktesios” .

29. Die sonderbare Wahl von Namen und Ubertragung von Epitheta beschränkt sich auch
nicht auf Homer. Die Redner der Ithakesierversammlung im ß (s.o. Abschn. 13) gehören ei-
nem viel späteren Dichter, die Benennung des Nestoriden Peisistratos in yéo zu Ehren der
Familie des Tyrannen erst dem letzten Bearbeiter. Umgekehrt hat schon vor Homer der
Dichter der Ur-Argonautika die Phäakenkönigin 'Apńrn wegen ihrer Rolle als Ehestifterin
zwischen Iason und Medeia nach Hera, der Góttin der Ehe, erfunden, ihr, weil Medeia sie
anfleht, diesen Namen 'Zu der man fleht' gegeben und noch dazu das stehende Beiwort der
Hera, AeukcoAevoc, auf sie übertragen. So ist sie in unsere Odyssee (f 303—315, n 53ff)
und, auf andern Wegen, in das Epos des Apollonios eingegangen? .

30. Ebenfalls vor Homer, aber als schon feststand, daß Achilleus höchsten Kriegsruhm ern-
tet und fern der Heimat einen frühen Tod findet im Dienst anderer, denen sein Verlust
schwere Niederlagen bringt, und als sein Name längst auf äxos bezogen wurde und sein
Charakterbild mitbestimmte, hat ein Dichter, spätestens 'Arktinos', in der Amazone ein
weibliches Gegenstück zu Achilleus geschaffen und mit Blick auf diesen benannt” ; die ge-
naue Entsprechung zu ‘Axt-A(A)-evs, mit mévdoc fiir dxos, *Iev3-A()-eta, metrisch un-
brauchbar bzw. unbequem, hat er nach Ilpwreoidaoc, ‘Apkeoidaoc, EK EOIMETAOS U.a. er-
weitert zu Ilevdeoikeıa.

31. Es haben also schon Homers Vorläufer, wie danach er selber und seine Bearbeiter, in
den Namen für neu erfundene Personen auf deren Vorbilder Bezug nehmen können. Sich
von solchen Gedankenassoziationen bestimmen zu lassen, gehörte offenbar zur Schaffens-
weise der frühen Dichter-Sänger überhaupt, und ebenso gewiß auch das Übernehmen von
Epitheta der Vorlage, unbesehen, ob sie zur neuen Person paßten. Es erklärt sich am ehe-
sten aus der jahrhundertelangen Mündlichkeit des vorgeschichtlichen Heldenliedes. Der
Übergang zur Schriftlichkeit ist ja nicht plótzlich erfolgt und war nie total, solange Sánger
auswendig vortrugen. Die beiden hier erórterten Besonderheiten epischen Erfindens, von
der doch intensiv betriebenen oral-poetry-Forschung unseres Jahrhunderts m.W. nicht er-
kannt, gewähren jedenfalls unerwartete und sichere Einsichten in die Schaffensweise der al-
ten Epiker und lassen die Überlegenheit Homers sichtbar werden.

7 Antike und Abendland 25 (1979) 141—147.
Pausanias 1, 14, 3.

7% Antike und Abendland 30 (1984) 148—153.

25 Antike und Abendland 25 (1979) 160ff.
Ein junger Mitforscher, David Bouvier, kommt unabhängig, auf Grund trefflicher Einzelbeobachtun-
gen, zum gleichen Schluß (briefliche Mitteilung).

Initial PIE *gWh- in Germanic

Edgar C. Polomé (University of Texas)

The probiem of the labiovelars and their reflexes in Germanic has recently been the object
of renewed interest in connection with the controversial hypothesis of Elmer Seebold
(1967; 1980) that the Indo-European labiovelar aspirate and the clusters *ghw and *ghw
(i.e. ‘velar’ or ‘palatal’ + -w-) become b- initially in Germanic, except before u where they
are reflected by g- Reviewing the evidence for labialization of labiovelars in Germanic,
Thomas Markey (1980) posits a rule establishing a major isogloss between East Germanic
and Northwest Germanic and accounting for the labialization of back spirants before
resonants, e.g. Go. bagms versus OHG boum ‘Tree’. For initial *g”h-, Markey focuses on
Gmc. *warma- ‘warm’ which he derives from IE *gWhor-mo-, but he sees a reflex of IE
*gWhr-mo- in ON Garmr, the name of the ‘hound of Hell’, which also appears in a kenning
for ‘fire’ (elris garmr). He accounts for NGmc. *garm- as follows:

IE *gWhr-mo-

PGmc. di: (delabialization before u from *r);
Gmc. *gorma- (lowering of u to o by a-Umlaut);
ON *garm- (late Gmc. coalescence of o and a).

The existence of such pairs as ON gandr vs. Da.dial. vann (OE wenn) < *g’hon(yo)- which
suggests “an apparently arbitrary alternation of g-/w- as reflexes initially before a” (Markey
1980: 287), would then imply similar diachronic explanations.

In a much more ambitious monograph on the Proto-Indo-European labiovelars A.G.E.
Speirs (1978) takes the occurrence of ‘irregular’ labials such as the initial b- of Latin bos
“ox, cow’ or the medial -f- in Gmc. *wulfa- ‘wolf’ as an excuse to question the whole tradi-
tional view about the treatment of the labiovelars in the Indo-European dialects. Listing
the apparent alternations between labial and non-labial reflexes of Indo-European gutturals
(e.g. Latin haedus: fedus [Sabellic] ‘young goat’; faba: haba [Faliscan] ‘bean’; OHG krioh-
han: OE créopan ‘creep’; Go. leihwan ‘lend’: OHG bi-liban ‘remain, be left’; etc.), Speirs
elaborates a theory ascribing most of these alternations as well as the ‘Reimwortbildungen’
of the same kind, described by H. Giintert (1914), to early Indo-European labialization and
dentalization processes, similar to those evidenced by Greek theino ‘strike, kill’: phonos
‘murder’ < PIE *gWhen/gWhon-. Accordingly, in his system SEW would become zt before
front vowel and *p before back vowel, being, however, delabialized to *k before u or as a
result of dissimilation processes. Thus, a Proto-Indo-European root *kWew- would appear
as *tew- ~*pow- ~*ku- under the conditions of the regular ‘ablaut’ alternations, but with

304

paradigmatic leveling and restructuring, this pattern would be likely to be reflected by the
three types: (1) tew- ~tow- ~tu-; (2) pew- ~pow- ~pu-; (3) kew- ~kow- ~ku- in the dia-
lects (Speirs 1980: 69). With such a hypothesis, Speirs is obviously able to account for the
most unexpected ‘correspondences’: since PIE *gWew-gW/h) ‘bend’ (Speirs 1980: 28) can
be reflected by *bhew-g-, *dhew-g(h)-, *dhew-b(h)-, etc., it can serve as the etymon of E
bow, duck, dip, to which — with the basic meaning of convexity — Speirs (1980: 31) adds
E cudgel, pock, cob ('round lump") and even pudding! As for cases where labiovelars really
do appear, e.g., Go. giman versus OE cuman ‘come’, Speirs (1980: 69) states: “The labio-
velar was no doubt restored in the third group [i.e. *kew- — *kow- — *ku-] before the
paradigm was irrevocably split."

It stands to reason that little credence can be given to a hypothesis that endeavors to derive
from a same root *gWher- in the expanded form *gWhr-em- Gk. chremizö ‘neigh, whinny’,
OCS griméti ‘thunder, roar’, Skt. bhramara- (bumble) bee’, OHG bremo ‘gad-fly’, OE
dream ‘joy, rejoicing, music, song’, because they all express various sounds, and then
proceeds to do the same with *g”r-em- (Gk. brémo ‘roar, clamor’), *ghel- (allegedly in
Gmc. *galan- ‘sing, call’ and *baljan- [OHG bellan] ‘bellow, bark’), *gel- (allegedly in ON
kalla ‘call, shout, say, tell’ and Lith. bylöti ‘say, speak’).! The gratuitous assumption of
extensive leveling and analogical restructuring is perhaps, beside the rather superficial
semantic analysis, the greatest flaw of this extremely speculative new approach to the
labiovelars.

The views of T. Markey are much closer to the traditional theory, whose major points he
usefully summarizes (1980: 287—288):

(1) Labiovelars never appear before u in the centum languages, where the labial element is
obviously redundant.

(2) Delabialization appears to be more extensive in peripheral centum languages like Old
Irish or Tocharian, but in Old Irish, at least, it must be a late development (post-
Ogamic, according to Markey, on account of the u of Olr. cruth ‘shape’ versus W pryd
‘time’ < *KWrtu-).

(3) The labial element does not appear in Germanic before Gmc. u, stops or resonants as
well as in final position.

(4) Labiovelars and clusters of velar + w are merged in Proto-Germanic and show the same
reflexes: /k¥/, /x¥ ~hW/, /gW ~yW/.?

(5) Intervocally, Gmc. /gW — yW/ appears to have been dissimilated to -g- before back
vowels (especially 4) and to -w- before front vowels, e.g. Go. magus ‘boy’ < *may"u-:
*may Wi 7» mawi ‘girl’. Analogical leveling has however distorted the original pattern,
especially in verbal paradigms.

As to initial TE *gh-, Markey (1980: 290) resolutely rejects the labial thesis of Seebold as
being based on specious etymologies like the connection of Go. bidjan with Gk. théssamai
‘pray’. For the etyma where IE *k™ is reflected dialectally by f, Markey (1980: 291)
points out that the complex dissimilation/assimilation processes involved in this residual
group of etyma (Go. fidwor ‘4’, twalif ‘12’; OF ri. fial ‘wheel’; ON ifr ‘liver’, etc.) “represent
a stage of late, dialect specific transformations of the labiovelars inherited by Germanic". It
is in the same perspective that he then tries to solve the problem of the aberrant cases with
Gmc. g- before -a- as reflex of IE *gWh-. While he agrees basically with the traditional rules:

305

(a) IE *gWh- before IE *o is reflected by *w- in Gmc., e.g. OE wearm: Skt. gharmäh
‘warm’;

(b) IE *gWh- before Gmc. *u is reflected by *g- in Gmc., e.g. ON gunnr, OE gup ‘battle’
(< Gmc. *gunpa-), Markey reorders the rules of a-Umlaut and lowering of o to a in
North-Germanic. This is precisely what makes his explanation of NGmc. *garm- in-
acceptable, because, if we applied the rules as implied by his hypothesis, the dia-
chronic development of PGmc. *burpa- would yield *barp instead of borp ‘farm-
house, village’ in Old Norse, according to the scheme:

PGmc. *burpa- (with ur from IE *r)
Gmc. *borpa- (a-Umiaut)

|
NGmc. *barp(a)- (lowering ofo to a).

It is, therefore, obvious that Markey's explanation of ON garmr as a reflex of IE *gWhrmo-
has to be rejected and can, accordingly, not serve as a model for the explanation of alterna-
tions like ON gandr ‘wand’ vs. Da. dial. vann ‘wen’.?

Markey (1980: 286), while discussing the views on the primordial existence Proto-Indo-
European labiovelars, clearly leans towards the hypothesis advocated by Jerzy Kuryłowicz
(1935: 1—26; 1956: 356—366) deriving the labiovelars from the velars through secondary
— conditioned — labialization. This hypothesis is analyzed in considerable detail by Lars
Steensland (1973), who reaches a different conclusion after a thorough review of the
available etymological material. He concludes (1973: 117) that the oldest system of Proto-
Indo-European must have had two ‘gutturals’ with the following realizations:

[kA : [kCOe, k*i, ko, ka]

/kB/ : [k¥e, kWi, kWo, kWa] E™

— but that neutralization before /o/ and /a/ occurred at a very early date, reducing the
system to:
[kA/ : [kKMe, kti]
/kB/ : [kWe, kWi]

[KA] is assumed to have a palatal articulation [kt], which led to its palatalization to [k'];
morphological leveling spread this [k'] before back vowels, especially o, less u, seldom a —
hence, secondarily [k'o, k'u, k'a] for [ko, ka, ku].

/kB/ was labialized, but with early neutralization of the labialization before [o] and [a] the
unlabialized forms were likely to spread before e, especially in roots with the ablaut eu~
ou~u — hence, secondary (k(*)e) for [kWe].^

[ko, ka, ku]

Steensland does not deal specifically with the reflexes of labiovelars in Germanic, but if he
is right in assuming that the series [kWe, kWi, ko, ku, ka] represent the “Ausgangspunkt für
die Gutturalsysteme aller urindogermanischen Sprachen” (1973: 117), it would imply that
what we reconstruct as *g”ho-, alternating with *g”he-, would actually already be dela-
bialized to [gho] in Pre-Germanic and accordingly be reflected by initial [ya] in Proto-
Germanic. Hence, ON geó (« PGmc. *yaója-; de Vries 1961: 159) ‘mind, mood, spirits’
would closely correspond to Olr. guide ‘prayer’ (< IE **Whodhyà; Pokorny 1959: 488)
and ON gerr, OHG garo, OE gearu (< PGmc. *yarwa-) ‘ready’ (lit. ‘cooked’ — cf. Du. gaar,
G. gar) could be derived straightly from PIE *g’hor-wo-, parallel to *gWhor-no- in Lat.

306

fornus ‘oven’, Olr. gorn ‘fire’ (Pokorny 1959: 494). But Steensland has obviously not made
up his mind about the reflexes of initial PIE *g”h- in Germanic, since he refers to See-
bold's rejection (1967: 112) of the above-mentioned etymology of ON ged and labels the
derivation of ON varmr from IE *g”’hormo- “umstritten” (1973: 14). However, he defi-
nitely assumes that initial g- corresponds to PIE *gWh- in OE gub (gyb in compounds)
‘combat’ < PIE *g’hntya, from the root *gWhen-{a-) hit’ (1973: 27).

This brief review of the recent discussion of the problem of the reflexes of initial *gWh- in
Germanic indicates that, while apparently no one wants to follow Seebold, no satisfactory
solution has yet been given either.

Summarizing the debate, it is, however, possible to make the following points:

(1) Wherever labials appear to reflect labiovelars in Germanic, each case has to be ex-
plained individually in its own dialectal and phonological context, as George Lane
(1939: 192) already indicated:

(a) with the numerals for ‘four’ and ‘five’ we are obviously dealing with assimilation
and analogical leveling processes, which could be tabulated as follows:

IE *KWetwòr- ‘4’ | *penkWe ‘5’

Pre-Germanic
Analogical leveling of anlaut *kWekW6r- *petwör | *pempe
Progressive assimilation [kW — tW] ^ [kW — kW] t [p - kW] - [p — p]
of second syllable anlaut 4 o Ä

Blending *pekWór- |
Proto-Germanic *fey Or *febwor *femfi
Gothic fidwor | fimf
Old Saxon fiuwar fif

(b) in the case of ON ofn, OHG ovan, OE ofen ‘oven, furnace’ versus Go. auhns,
OSwed. ugn, the possibility of a Celtic loan alongside the inherited term is to be
considered in view of the more advanced metallurgic technology of the Celts
(Birkhan 1970: 142-147);

(c) in the case of the ‘wolf’ the occurrence of -p- forms in both Germanic (Go. wulfs)
and Italic (Lat. lupus) suggests a Pre-Germanic blending of IE *w/kWos (meta-
thetized *4kWos) with the root *wlp- of Lat. volpés ‘fox’, Lith. vilpi$ys *wildcat'
(Lane, 1939: 194),° perhaps ascribable to tabuization.

(2) The etymologies advanced by Seebold (1980: 477—484) to assume the labialization of
initial PIE *gWh- in Germanic are unconvincing:

(a) ON bani, OE bana, OHG bana ‘killer’ would undoubtedly fit semantically under
PIE *gWhen- ‘hit, slay’, but the connection with Av. banta- ‘ill’ is unimpeachable:
illness is due to the action of evil spirits in Zoroastrian belief; they strike the
victim like the killer inflicts a mortal wound (OS beni-wunda). From the same
root *bhen- 'hit, injure' (Pokorny 1959: 126) Germanic also derives MHG bane
(G Bahn), MDu. bane (Du. baan) ‘open space, road’, which points to an original
meaning ‘beaten ground, flattened area’ (de Vries 1971: 24).

(b) Go. brinnan, ON brinna, OE beornan, OHG brinnan “burn”, allegedly a -nu-present
to *gWher- ‘burn’ (*gWhr-nw-; cf. Skt. ghrnoti ‘glows, gleams’, Arm. jernum ‘I
heat, burn’) with an analogical full grade vowel — Gmc. *brenn- instead of brunn- —

(c)

307

has been connected with the theme (IT) *bhr-ew- of the root *bher- ‘gush’ or ‘boil
up’ (of water), ‘flare up’ (of fire) (Pokorny 1959: 132—133, 143—145), express-
ing the concept of *welling forth' in Gmc. *bruno, *brun(e)n- *spring' (Go. brunna,
OE burna, OHG brunno) and the concept of ‘leaping up flames’ in Gmc. *brinnan-
(with its causative *brannjan- in Go. brannjan, ON brenna, OE berna, OHG
brennan). Rather than the infixation of a nasal in the theme (II) *bhr-ew-, it
might, however, be preferable to assume a theme (II) *bhr-en-, followed by a
present tense suffix -w-, expanded over the whole paradigm (hence -nn- < *-nw-)
and the derivatives (van Coetsem 1972: 193): this would account for the -n- —-nn-
alternation in forms like ON bruni ‘burning, fire”: brenna ‘burn’ and be paralleled
by the theme *bhr-em- in Olnd. (RV) bhramah ‘whirling flame’, bhrmih ‘restless,
active’ (applying to Agni, RV I 31.16). As for the semantic development (‘flare
up’ > ‘burn’), the connotation ‘licking flames’ is further substantiated by related
terms such as the Vedic intensive jérbhuriti or Mir. breo (< *bhriwo-) ‘flame’.

The derivation of Go. bidjan, ON bivja, OE biddan, OHG bitten ‘ask, entreat,
demand’ from *g’hedh- ‘ask, demand’ (Pokorny 1959: 488) is described as “an
sich bestechend" by Wolfgang Krause (1968: 123—124). Seebold's comments on
the use of Ved. badh- (1980: 480—481) are, however, extremely lopsided, focus-
ing only on a specific use of the term in the Vedic compound jfu-badh- describing
a devotional genuflection: pointing out that the corresponding use of OS an
kneobeda and its parallels (OE on cneowgebedum, ON d knébe&u) reflects an
effort to translate into Germanic this Oriental practice, he concludes rather hastily
that "die angeführte etymologische Gleichung nur einen Sinn hat wenn das germa-
nische Wort ‘Kniebeuge’ bedeutete" (1980: 482). This is, however, disregarding
the extensive use of RV badh- in other contexts where it essentially seems to
mean ‘to urge’, e.g. in RV I 61,2, where Geldner (1951: 1.78) considers badhe as
an infinitive which he translates tentatively as ‘nötigen’ (a song of praise is offered
to Indra to soothe and impel him); similarly, in RV I 132, 5, badhe refers to
efforts to ‘overwhelm’ the deity with praise; in RV VI 11, 1, the sacrificing priest
is called upon to act as ‘urged on’ by the Maruts (Geldner 1951: 2.103); etc. This
concept of ‘pressing, forcing’ tallies with the meanings ‘order, require’ appearing
in early southern Germanic sources, e.g., OHG pitit. poscit (Ahd. Gl. I. 230. 14),
OE ic bidde. peto (AElfric), beside ‘pray, request, entreat’, and as Alfred Goetze
(1939: 344) rightly points out: “bitten kann dauernd den verschiedenen Haltun-
gen von ‘befehlen’ bis ‘flehen’ gerecht werden, je nach dem gesellschaftlichen Ver-
haltnis des Sprechenden zum Angesprochenen”. Since PIE *bhedh- (Pokorny
1959: 114) is formally unimpeachable and semantically quite acceptable as the
source of Gmc. *bidjan, there is no need to look elsewhere for etymological linkage.

The other etymologies adduced by Seebold (1980: 482—484), which include
such improbable suggestions as the derivation of Gmc. *bero(n) *bear' from IE
*ghwer- ‘wild animal’ (Gk. ther, Lat. ferus, OCS zv£ri, etc.) or the assumption of a
totally unattested base *ghwa- (from *ghewa- ‘invoke, entreat’ — listed by Po-
komy [1959: 413] as *ghaue- ‘rufen, anrufen’) to account for Gmc. bö-ni- ‘re-
quest, prayer’ (ON bon, OE ben), hardly need to be discussed.

No irrefutable cogent argument can be derived from the evidence adduced to in-
dicate that Gmc. b- reflects initial PIE *gWh- (&hw-).

308

(3) Initial PIE *gWh- is delabialized before u and before resonants, as indicated by ON
guór, gunnr, OS güdea, OE gub ‘combat’, OHG gundfano ‘battle-flag’, reflecting Gmc.
*gunb- from PIE *gWhnt-.®
Unfortunately, the other evidence is rather weak, consisting mainly of Germano-Greek
isoglosses, in which the Greek component is merely a lexical entry in Hesychius’
glossary, e.g., MHG gumpen ‘jump, hop, dance’: Gk. athembotisa. akolastainousa
(boisterous’)’ — or — Go. gund ‘gangrene’, OE gund, OHG gunt ‘pus’: Gk. kanthule
‘tumor’ (cf. fn. 3, supra). ON grunr ‘suspicion’: Gk. phrén ‘midriff? (as seat of the emo-
tions), phronéd ‘think’ is presumably to be rejected, as Gk. phrén seems to be linked
with phrazö ‘point out, explain’ within Greek (Chantraine 1980: 1225, 1228).

(4) PIE *gWh- has been assumed to be delabialized before PIE *o in:

(a) ON ged ‘wits, disposition’ < PIE *gWhodhyom (cf. Olr. guidim ‘I pray’, Gk.

pôthos ‘longing’);

(b) ON gandr “magic staff’ < PIE *gWhondhos (from *gWhen- ‘strike, slay’);

with *gWhmb- in MHG gumpen).

None of these examples are, however, convincing:

(a) For ON geö, Seebold (1980: 459--460) made a fairly good case to establish that
the basic meaning of the term was ‘that which is fitting’, which is precisely what
George Lane had already suggested (1936: 23) to connect the term with OHG
gigat ‘fitting’, OS gigado his like’, from the root in OFris. gadia ‘unite’, MLG
gaden ‘fit, please, pair’. ON geó would then reflect Gmc. *yaója- « PIE *ghodhyo-
(cf. OCS godü ‘right time’, god& byti ‘fit, please’, godinù ‘pleasing’; Pokorny
1959: 423-424), with a shift in meaning from ‘what is appropriate for someone’
to his ‘characteristic disposition’.

As indicated in footnote 3 (supra), the original meaning of gandr focuses on the
concept of magic and the etymology of the term is therefore to be sought along
that line (de Vries 1957: 362).

‘exuberant’ apply to MHG gampen as well.

As for the new example suggested by Markey (1980: 292), the name of hound of hell, ON
Garmr, has been quite attractively explained in its mythological context by Bruce Lincoln
(1979: 276—285) as a derivation from the onomatopoeic root *gher- (Pokorny 1959: 439).

If, then, there is no cogent evidence of delabialization of PIE *gh- before PIE *o, should
we assume that, in initial position, PIE *g”ho- developed into Gmc. *wa- on the basis of
Gmc. *warma- ‘warm’: Lat. formus, Av. garama- ‘hot’, Skt. gharma- ‘heat, warmth’, etc.?

N”

Several of the etymologies adduced to back up this assumption (e.g., Polomé 1948) should
be abandoned, e.g., the derivation of Go. wandus, ON vondr ‘rod, twig” from PIE *gWhen-
‘strike’;® the reconstruction of a PIE *gWhel- ‘yellow’ beside *ghel- (Pokorny 1959: 429)
to account for Du. wouw ‘reseda luteola’ and Lat. fel ‘spleen’ (cf. Seebold 1980: 468—
469); the comparison of OHG wahs ‘sharp’ with Gk. phoxos ‘pointed’ (cf. Seebold 1980:
473—474). The case of Go. wamba ‘belly’: Olnd. gabhah ‘vulva’ is more complex: the
effort of Seebold (1980: 470-471) to assign the meaning ‘groin’ (G. ‘Schenkelspreize’) to
gabhd- is rather unconvincing,” but even if the meaning ‘vulva’ is maintained for Skt.
gabhah, it does not support linkage with Gmc. *wambo ‘belly, abdomen’. If the Old Indic

309

form has a labiovelar initial, it would rather be *gW- in PIE *gWembh-/ *gWmbh- ‘plunge,
sink down’, also appearing in Gk. bdprò ‘dip’, baphé ‘dipping’ — without nasal infix: ON
kaf ‘depth of the sea; diving’, OSwed. kvaf; ON kvefia ‘suffocate’ (Mayrhofer 1956: 1.324;
Pokorny 1959: 465—6; de Vries 1961: 296, 336; Chantraine 1968: 156, 164). That leaves
the Germanic terms (Go. wamba, ON vomb, OE womb, OHG wamba) fairly isolated, be-
cause the alternate connection with Skt. vapd ‘omentum, skin investing the intestines’, is
improbable (Polomé 1948: 565; Skt. vapà rather belongs to vépati ‘shears, shaves’ [terms
for ‘skin’ are frequently deriving from roots meaning ‘cut’; cf. de Witte 1950: 60ff.], as
Mayrhofer 1976: 3.144 points out).

A direct connection between Skt. gadhah ‘swallow’, gadhdm ‘ford, shoal’ and Gmc. *wada-
‘ford’ (ON vada, OE wadan, OHG watan ‘wade’) is fairly well excluded by the latter’s
correspondence with Lat. vadum ‘ford’: the Old Indic term seems to be an old ‘thyme-
word’ to PIE *wadhom “ford’ (Mayrhofer 1956: 1.334).

This practically leaves Gmc. *warma- as the only cogent example of the derivation of initial
Gmc. *wa- from PIE *gVo-, but there is the alternate etymology deriving the term from
the root *wer- *bwen' (Hitt. uar- 'burn', OCS varü "heat", etc A.

The parallel derivation of Gmc. *wö- from PIE *gWhG- is also based on one isogloss: Go.
(dauns) wopi ‘sweet (smell), ON gòri ‘better’, OE weóe ‘sweet, pleasant’, OHG wuodi
‘sweet, mild’: Gk. phötion. prosphiles, hédi (Hesychius)!° — obviously rather weak evi-
dence to reconstruct a PIE *gWhotyo- (or *ghwotyo-) since there is no other trace within
Greek of *phot- (— *thet-) with the meaning 'sweet'. On the other hand, the alternate con-
nection of the Germanic terms with Olnd. (RV) vara- 'dear, fair, noble' (de Vries 1961:
684) is unconvincing (‘wertlose Wurzeletymologie’, according to Mayrhofer 1976: 189).

Consequently, lack of sufficiently valid evidence compels us to conclude that the treatment
of initial PIE *g“h- before mid-level back vowel (*6) in Germanic remains an unsolved
problem: labialization to b- is improbable; delabialization to g- is unwarranted; assumed
development to w- rests on disputable etymologies.

(5) The treatment of PIE *gWh- before front vowel in Germanic is even more difficult to
assess, since there are no clear etymologies illustrating this development: from the paral-
lelism of *gW- and *kW-, the maintenance of the labiovelar would be expected, but there is
no evidence of initial Gmc. [yW] from *gWh- comparable to [kW] in Go. giman ‘come’ from
PIE *gWem- or [xW] in Go. hveila ‘time’ from PIE *kWey- (cf. Lat. quiés ‘rest’ [*KWye-]:
ON hvid ‘rest”). Labialization to b- is fairly well excluded; assumed development to w- rests
on unacceptable etymologies such as:

(a) OHG folle-wemon. fluitare (‘swarm’, Notker): Russ. gómon ‘din, clamor’, alle-
gedly from PIE *gWhem- ‘swarm’ (Polomé 1948: 567). The Germanic terms, which
also include OS wemmanthi. scaturiens (‘gushing out |of blood], Prudentius glosses),
MHG wimmen ‘move, swarm’, OHG wimidon, wimizzen ‘bubble up, swarm’, Du.
wemelen ‘swarm’, etc. (de Vries 1971: 827), belong presumably to the hardly ety-
mologizable affective/onomatopoeic vocabulary, like wabble, wobble, and the like
(Seebold 1980: 472).

(b) Go. wairs ‘worse’, ON verr, OE wiers, OHG wirs: OCS goriji ‘worse’, with an
assumed original meaning ‘(more) bitter’, from PIE *g”her- ‘burn’ (Wood 1926: 103).
Since OCS goriji rather belongs with Gk. A/iciron ‘worse’ (< *gher-yes-; cf. Stawski

310

1956: 321; Machek 1968: 176—177; Seebold 1980: 472), the Germanic terms have to
be explained otherwise — perhaps, as Elof Hellquist (1948: 1398) suggested, from IE
*wer-s- high, upper’ (cf. Skt. värsiyas- ‘higher’, Lith. virus ‘upper part, pinnacle’;
Pokorny 1959: 1152) with the meaning ‘over and above the right measure’ (cf. OE
yfel ‘evil’ [< *upé-lo-] versus the superlative ufemest, yfemest ‘highest, uppermost’
[= Skt. upamd-]).!!

(c) Go. waila wisan ‘revel’, OE wesa (Gl.) ‘intemperate drinker’, OHG firwesan ‘con-
sume’: Skt. ghdsati ‘consumes, devours, eats’, allegedly from PIE *gWhes- Though the
separation of the Germanic terms leaves Skt. ghdsati practically without etymology
(cf. Mayrhofer 1956: 1.358; 1976: 3.699), they definitely belong with OInd. (RV VIII
4.8) anu vavase ‘has devoured’, Avest. vastrom ‘fodder’, Mir. fess, feiss (*wes-ta)
‘meal’, W gwest ‘feast’, ON vist, OHG wist ‘food’ under PIE *wes- ‘revel’ (Feist 1939:
568; Pokorny 1959: 1171; de Vries 1961: 669). No valid answer can accordingly be
given to the question of the development of PIE *g'’A- initially before mid-level front
vowel (*e) in Germanic.

Notes

Thus, the terms listed by Pokorny (1959: 203—204) under *der- (2) und *ger- (2) (1959: 383—385)
are lumped together with those appearing under "gWer(a)- (1959: 478). To be sure, as Pokorny
(1959: 478) points out: “Gegenüber den ähnlichen Schallwurzeln ger-, $r- ist nicht überall eine
sichere Scheidung möglich”, but a semantic difference is clearly recognizable ("für die ... unter
gler- vereinigten Worte [ist] der Begriff der gehobenen Äußerung unverkennbar, bis auf die darum
nicht sich anzugliedernde dh- Erweiterung" [e.g. Lit. ge£das 'shouting']). On the other hand, the re-
construction of a root *gWhel- "expressing various types of sound” (Speirs 1978: 46) ist strictly
arbitrary: there is no direct evidence of the labiovelar, and it is assumed on the basis of a delabialized
form *ghol- reflected by Gmc. *galan- (OHG galan ‘sing, call, resound, cry’, nahtigala ‘nightingale’,
etc.) — with -dh- suffix in ON galdr ‘song, charm spell, sorcery’ -- and of a labialized form *bhel-
(Lith. bélsti ‘make a noise by banging or knocking’, MHG boldern ‘make a din’, etc.) — with ex-
panded forms such as *bhel-s- in Skt. bhdsd ‘speech, language’, Lith. balsas ‘voice’, etc. and *bhl-ew-
in Lith. bliduti ‘bleat, blubber, howl’. Again, a number of apparently unrelated roots (*ghel- (Pokor-
ny 1959: 428), *bhel- (6) (Pokorny 1959: 123—124); *bhleu- (Pokorny 1959: 158); etc.) are linked
together with the flimsy excuse that “in words denoting sounds some have found a happy hunting
ground for onomatopoeia”. Wondering “how it is known what was onomatopoeic to the ears of
Indo-European communities more than five thousand years ago”. Speirs (1978: 46) wishes “to draw
attention to the fact that the initial consonants, as reflexes of what we believe to have been labio-
velars, follow the same pattern. . .". He could, of course, have done his home-work a bit better and
also brought nicely together under labiovelar roots the set of forms meaning ‘sticking out’; listed by
Pokorny under *gher- (3) (1959: 440) and *bher-em- (1959: 142), e.g. G Granne ‘awn’ and Brom-
beere ‘blackberry’ (< IE *“bhré-mo- ‘thorny [shrub]’; cf. OE bròm ‘broom, furze’); E brim, etc., or
added to the alleged reflexes of PLE *kWer-kW- (Speirs 1978: 52) OHG felga, OE fielg ‘felly? < IE
*nel-k- (Pokorny 1959: 807). But pursuing Speirs’ reasoning ad absurdum, there would be few roots
left with original labials, dentals or velars, while labiovelars would astonishingly proliferate, and one
wonders with his promised revision of the current ablaut theory (Speirs 1978: 69) what will be left
of the generally accepted Proto-Indo-European phonological system?

Cf. Moulton (1972: 171): The voiced labiovelar /gW ^'yW/ "seems to have been a unit phoneme in
Pre-Gmc, but not in PGmc”. The voiceless labiovelars /kW/ and /xW —hW/ “may well have been the
consonant clusters /kw/ and /xw ~hw/ rather than unit phonemes”. Actually, the consonant shift
yielded [xW] from IE *kW, except under the conditions of Verner’s Law when [yW] resulted, but
[xW] became gradually [hW] in various environments in the different Germanic dialects (where their
distribution parallels that of /x ^h/ from IE *X; Moulton, 1972: 173).

311

7 It is not clear whether Markey (1980: 292) assumes any direct relationship between these two terms,
but obviously semantic considerations would rather exclude it. Wood (1926: 106-107) connected
Da. dial. vann, OE wenn ‘wen’, MDu. wan ‘tumor’, with OHG gunt ‘pus’, Go. gundr ‘gangrene’, and
derived them all from IE *gWhen- ‘swell’. As Seebold (1980: 462) points out, the basic meaning of
the root *gWhen- with which Skt. ghaná- ‘hard, firm, thick’, Lith. gand ‘enough’, Gk. euthenös
‘thriving, flourishing’ are usually connected, is rather ‘coagulate, curdle' (hence ‘thick, compact’, and
‘thrive’, respectively), which does not strengthen his semantic argument against the derivation of the
Germanic term for ‘pus’ from this root. The alternate link with the Hesychius-gloss kanthulé ‘swelling
tumor’ remains disputable, and other explanations are plausible for the Greek term (Chantraine
1970: 492). As for NWGmc. *wanja- > OE wenn ‘wen’, MLG wene ‘tumor, wart’, etc., it has also
been connected with OHG wanast, wenist ‘ramen’, which is not more convincing! ON gandr, on the
other hand, should be glossed ‘monster, magic’, according to Seebold (1980: 458), who rejects the
meaning ‘wand’, referring in a footnote to Elof Hellquist (1893), which he could not consult. This
information, like that of most etymological dictionaries, is presumably based on Sveinbjórn Egilsson,
who states (1931: 170) that ‘wand’ is the basic meaning of the term, referring specifically to Hell-
quist (1893) and identifying gandr with volr, a round thick staff described as ‘redskab til udgvelse af
völvers virksomhed’. This magic staff is used by fortunetellers and then called spégandr (cf. Voluspa,
st. 29: fekk spjoll... ok spa ganda); the phrase vitta ganda means ‘cast a spell’, and in Voluspá, st.
22, it applies to Heiór plying witchcraft. Confirmation of the meaning ‘staff’ is sought in the com-
pound gandreiò ‘ridt pà en stok', but Jonsson (1931: 170) has to acknowledge that in the kenning
skjalda gandreid gandr has to be interpreted as a ‘harmful being’, i.e. the shield’s wolf. This ties in
with the fact that the wolf is the witches’ mount and has led Hugo Gering (1927: 28~29) to the con-
clusion that ‘wolf’? must be the older meaning, from an original etymon designating a magic being
(‘Zauberwesen’) and applying secondarily to objects such as a magic wand. Obviously such an as-
sumption makes derivation from IE *gWhen- ‘hit, strike, kill’ less plausible. It is, however, preferable
to consider ‘magic object’ as the original concept, in which case the connection with ON ginn- ‘out-
standing’, e.g. in ginnregin ‘supreme gods’, ginna ‘fool, delude, practice magic’, Ginnarr (= Ovinn),
lit. ‘magician’, Run. ginArunAR (7th c.) ‘magical powers’ (de Vries 1930) would deserve further
consideration (cf. de Vries 1961: 155, 167—168).

Though Markey (1980: 286) does not refer to Steensland (1973), his remarks about the thesis postu-
lating primordial palatals and labiovelars at the exclusion of a separate velar series are quite relevant:

"Structurally speaking alone, the velars are a necessary component in the opposition between palatals
on the one hand and labiovelars on the other: their elimination prohibits realization of the palatal
quality of the palatals and the labial quality of the labiovelars”.

For a review of Steensland's work, cf. Adolf Erhart (1974: 68—69).

OBret., Corn. louuern ‘fox’ reflects *lowerno- and is not related (Campanile 1974: 74).

For a thorough analysis of the phonological and morphological problems involved, cf. Seebold
(1980: 451—457).

Neither Frisk (1960), nor Chantraine (1968) lists Gk. (Hesych.) athemboüsa, which does not appear
in Liddell-Scott (1940—1968) either. As is well-known, in Hesychius many glosses without ethnic
reference or specific localization are not actually Greek.

Rather than following Seebold (1980: 466) in his acceptance of Jost Trier's connection with Olnd.
vanas-, interpreted as ‘cluster of foliage’ (G. ‘Laubbiischel’) in RV X 172.1, and further with Skt.
vdna- ‘forest, tree’, we would give preference to the traditional derivation from IE *wendh- ‘turn,
wind, twine’ (Pokorny 1959: 1148), as the term corresponds to OHG want ‘wall’ — made of inter-
twined twigs plastered with loam (cf. ON vandahis, lit. ‘wickerwork-house’); this etymology is
further confirmed by such terms as Swed. dial. vann ‘tendri? (of pea plant).

In the passage quoted from the Vajasaneyi Samhita XIII. 24, there is no cogent reason to reject the
meaning ‘penis’ given to musti- (Mahidhara; cf. Monier-Williams 1899: 824b): the term is related to

muskah (RV) ‘testicle’, muskdu (AV) ‘pudenda muliebra’ (Mayrhofer 1963: 2.657—-658) and

belongs to the common semantic transfer of ‘mouse’ to parts of the body (e.g. Lat. musculus ‘muscle’;
cf. De Witte 1950: 67, 350); the sexual connotations of those derivations have been demonstrated
by Hermann Giintert (1913), and parallels like MHG miuselin ‘membrum virile’ ~ ‘little mouse’, as
weil as the symbolism of obscene gestures with the clenched fist (e.g., in the Mediterranean area)
illustrate the fact that the coexistence of meanings like ‘penis’ and ‘fist’ on the basis of their *Mause-

312

schaft (‘mousedom’; Thieme 1951: 214, fn. i) is not excluded. Skt. gabhe mustim atamsayat ac-
cordingly refers to the act of copulation and gabhá- is the ‘vulva’, literally: the ‘hollow’, the ‘recess’,
since gabhá- cannot be separated from gámbha- 'depth', gambhirá- ‘deep’, Avest. jafra- *deep', JafnuX
‘hollow, recess’. That this isa common way to designate the female genitalia is abundantly illustrated
by the relevant terminology (De Witte 1950: 272ff.).
The doubts expressed by the 1862 editor of Hesychius’ lexicon, quoted by Seebold (1980: 472,
fn. 92) are apparently not shared by Liddell and Scott (1940: 1969), but neither Frisk (1970), nor
S Chantraine (1980) lists phôtion.
Cf., e.g. De Vries (1961: 657).
The derivation from *ayer- ‘lower’, proposed by Seebold (1980: 472), is unconvincing: Skt. avarah
‘inferior’ is the comparative of áva ‘down’ versus the superlative (RV) avamd- ‘undermost, lowest’,
and RY avár ‘down’ is a locative adverb in -r; there is accordingly no ground to reconstruct a Proto-
Indo-European theme *H2w-er-, enlarged by -s- in Germanic, on the basis of such evidence.

References

Bennett, William H., 1969. “Pre-Germanic /p/ for Indo European /kW/”, Language 45: 243-247.

Birkhan, Helmut, 1970. Germanen und Kelten bis zum Ausgang der Römerzeit. Der Aussagewert von
Wörtern und Sachen für die frühesten keltisch-germanischen Kulturbeziehungen (= Österreichische
Akademie der Wissenschaften, philologisch-historische Klasse, Sitzungsberichte 272) (Vienna: Her-
mann Böhlaus Nachfolger).

Campanile, Enrico, 1974. Profilo etimologico del cornico antko (= Biblioteca dell’Italia Dialettale e di
Studi e Saggi Linguistici, vol. 7) (Pisa: Pacini).

Chantraine, Pierre, 1968. Dictionnaire étymologique de la langue grecque. Histoire des mots 1, A — A
(Paris: Editions Klincksieck).

Chantraine, Pierre, 1970. Dictionnaire étymologique de la langue grecque 2, E — K (Paris: Editions
Klincksieck). |

Chantraine, Pierre, 1974. Dictionnaire étymologique de la langue grecque 3, ^ — I (Paris: Editions
Klincksieck).

Chantraine, Pierre, 1977. Dictionnaire étymologique de la langue grecque 4a, P — Y (Paris: Editions
Klincksieck).

Chantraine, Pierre, 1980. Dictionnaire étymologique de la langue grecque 4b, b ~ N (Paris: Editions
Klincksieck).

van Coetsem, Frans, 1972. “Proto-Germanic Morphophonemics”, Toward a Grammar of Proto-Ger-
manic edited by van Coetsem, Frans, and Kufner, Herbert L., 175—209 (Tübingen: Max Niemeyer).

Egilsson, Sveinbjörn, 1931. Lexicon Poeticum Antiquae Linguae Septentrionalis/Ordbog over det Norsk-
Islandske Skjaldesprog, 2nd edition by Vinnur Jónsson (København, S.L. Møllers Bøgtrykkeri). —

Erhart, Adolf, 1974. Review of Steensland, “Die Distribution der urindogermanischen sogenannten
Gutturale", Slavia 43: 68-69.

Feist, Sigmund, 1939. Vergleichendes Wörterbuch der gotischen Sprache mit Einschluß des Krimgoti-
schen und sonstiger zerstreuter Überreste des Gotischen (3rd edition) (Leiden: E.J. Brill).

Frisk, Hjalmar, 1960. Griechisches etymologisches Wörterbuch 1, A — Ko (Heidelberg: Carl Winter).

Frisk, Hjalmar, 1970. Griechisches etymologisches Wörterbuch 2, Kı — N (Heidelberg: Carl Winter).

Frisk, Hjalmar, 1972. Nachtrdge — Wortregister — Corrigenda (Heidelberg: Carl Winter).

Geldner, Karl Friedrich, 1951. Der Rig-Veda, aus dem Sanskrit ins Deutsche übersetzt und mit einem lau-
fenden Kommentar versehen, 4 volumes (= Harvard Oriental Series, volumes 33-36) (Cambridge,
Mass.: Harvard University Press).

Gering, Hugo, 1927. Kommentar zu Liedern der Edda 1. Götterlieder (Halle (Saale): Buchhandlung des
Waisenhauses).

Goethe, Alfred (ed.), 1939. Trübners Deutsches Wörterbuch, Vol. 1: A -- B (Berlin: Walter de Gruyter).

Güntert, Hermann, 1913. “Ar. musti- “Faust” und die Hexe MüS”, Zeitschrift für vergleichende Sprach-
forschung 45: 193—204.

Güntert, Hermann, 1914. Über Reimwortbildungen im Arischen und Altgriechischen. Eine sprachwis-
senschaftliche Untersuchung (Heidelberg: Carl Winter).

313

Hellquist, Elof, 1893. Etymologische Bemerkungen (Gefle).

Hellquist, Elof, 1948. Svensk etymologisk Ordbok, 3rd edition, 2 volumes (Lund: C.W.K. Gleerup).

Krause, Wolfgang, 1968. Handbuch des Gotischen, 3rd edition (Munich: C.H. Beck).

Kurytowicz, Jerzy, 1935. Etudes indo-europeennes 1 (Polska Akademja Umiejetnosci — Prace Komisji
Jezykowej, Nr. 21) (Kraköw: Gebethner and Wolff).

Kurytowicz, Jerzy, 1956. L ‘apophonie en indo-européen (Polska Akademia Nauk -. Prace Jezykoznawcze,
Nr. 9) (Wroclaw: Zaktad Imienia Ossolinskich).

Lane, George S., 1936. “The Labiovelars before 6 in Germanic”, The Journal of English and Germanic
Philology 35: 17—26.

Lane, George S., 1939. “Pure Labials from Labiovelars in Germanic”, The Journal of English and Ger-
manic Philology 38: 184—200.

Liddell, Henry George/Robert Scott, 1940. A Greek-English Lexicon. New Edition, by Henry Stuart
Jones, 2 volumes, 1 a — kwy, 2 à — wwöns (Oxford: Clarendon Press).

Liddell, Henry George / Robert Scott, 1958. A Greek-English Lexicon. Supplement, by E.A. Barber (Ox-
ford: Clarendon Press).

Lincoln, Bruce, 1979. “The Hellhound”’, Journal of Indo-European Studies 7: 273—285. 5

Machek, Vaclav, 1968. Etymologický slovník jazyka českého, 2nd revised edition (Praha: Ceskoslovenské
Akademie V&d).

Markey, Thomas L., 1980. “Delabialization in Germanic”, Folia Linguistica Historica 1/2: 285--293
[An earlier version of the same article appeared in the Journal of Indo-European Studies 8 (1980):
199—211.]

Mayrhofer, Manfred, 1956. Kurzgefaßtes etymologisches Wörterbuch des Altindischen 1 A — TH (Hei-
delberg: Carl Winter).

Mayrhofer, Manfred, 1963. Kurzgefaßtes etymologisches Wörterbuch des Altindischen 2 D — M (Hei-
delberg: Carl Winter).

Mayrhofer, Manfred, 1976. Kurzgefaßtes etymologisches Wörterbuch des Altindischen 3 Y — H; Nach-
träge und Berichtigungen (Heidelberg: Carl Winter).

Mayrhofer, Manfred, 1980. Kurzgefaßtes etymologisches Wörterbuch des Altindischen 4 Register (Hei-
delberg: Carl Winter).

Monier-Williams, Monier, 1899. A Sanskrit-English Dictionary, etymologically and philologically ar-
ranged (New edition, with the collaboration of E. Leumann and C. Cappeller) (Oxford: Clarendon
Press).

Moulton, William G., 1972. “The Proto-Germanic non-syllabics (consonants)", Toward a Grammar of
Proto-Germanic, edited by Van Coetsem (Frans) and Kufner (Herbert L.),141—173 (Tübingen: Max
Niemeyer).

Pokorny, Julius, 1959. Indogermanisches etymologisches Wörterbuch (Beın: A. Francke).

Polomé, Edgar C., 1948. “Over de etymologie van Nederlands wouw ‘reseda luteola’”, Belgisch Tijd-
schrift voor Philologie en Geschiedenis 26: 541-568.

Seebold, Elmer, 1967. “Die Vertretung von idg. g4h im Germanischen”, Zeitschrift für vergleichende
Sprachforschung 81: 104—133.

Seebold, Elmer, 1980. “Etymologie und Lautgesetz”, Lautgeschichte und Etymologie, edited by Mayr-
hofer, Manfred e£ al. (Akten der VI. Fachtagung der Indogermanischen Gesellschaft, Wien, 1978),
440—484 (Wiesbaden: Ludwig Reichert Verlag).

Stawski, Franciszek, 1956. Stownik etymologiczny jezyka polskiego, Vol. 1 (Krak6w: Naktad Towar-
zystwa Mitosniköw Jezyka Polskiego).

Speirs, A.G.E., 1978. The Proto-Indo-European Labiovelars (= Bibliotheca Slavica vol. 18) (Amsterdam:
Adolf M. Hakkert).

Steensland, Lars, 1973. Die Distribution der urindogermanischen sogenannten Gutturale (= Acta Uni-
versitatis Upsaliensis — Studia Slavica Upsaliensia Vol. 12) (Uppsala).

Thieme, Paul, 1951. “Der Lachs in Indien", Zeitschrift für vergleichende Sprachforschung 69: 209—216.

de Vries, Jan, 1930. “Ginnungagap”, Acta Philologica Scandinavica 5: 41—66.

de Vries, Jan, 1957. Altgermanische Religionsgeschichte 2. Die Götter — Vorstellungen über den Kos-
mos — Der Untergang des Heidentums, 2nd edition (Berlin: Walter de Gruyter).

de Vries, Jan, 1961. Altnordisches etymologisches Wörterbuch (Leiden: E.J. Brill).

de Vries, Jan, 1971. Nederlands etymologisch woordenboek (Leiden: E.J. Brill).

De Witte, A.J.J., 1950. De betekeniswereld van het lichaam. Taal-psychologische, taalvergelijkende
studie (Utrecht: Het Spectrum/Antwerp: Standaard Boekhandel).

Wood, Francis A., 1926. Postconsonantal w in Indo-European (= Language Monograph No. 3) (Phila-
delphia: Linguistic Society of America).

All our ‘yesterdays’

Jaan Puhvel (University of California, Los Angeles)

In a short monograph of his later years Karl Brugmann (1917) pointed out that Indo-
European possesses a common etymon for ‘yesterday’ but not for ‘today’ or ‘tomorrow’;
for the latter, parallel transparent means of expression predominate, as is still the case with
the English terms. Thus ‘today’ is typically made up of some demonstrative pronominal
element and a word for ‘day’, as in Skt. a-dyä, Hitt. anisiwat(ti) or kedani siwatti, Arm.
aysawr, Gk. ońuepov (< *xt-auepov)', Alban. sot (< *kya-diti), Lith. Yiafi-dien?, OCS
dini-si, Russian segö-dnja, Irish in-diu, Welsh he-ddyw, Lat. hodie (Faliscan foied), Gothic
himma daga, OSax. hiu-diga, OHG hiutu? , ON i-dag. Equally consistently, ‘tomorrow’ is
derived from or cognate with ‘morning’, e.g. Skt. svds (cf. Avest. stiram ‘early in the morn-
ing’), Hitt. lukkatti (siwatti) or lukkatta (siwatti) (lukat- ‘morning’), Arm. vatiy (val ‘early’),
Gk. atpwv (*awsr- ‘dawn’, as in Lith. au%ra, cognate with nus*), Lith. rytdj (rytas ‘morn-
ing’), Russian za-vtra (utro ‘morning’; cf. OCS za ustra, cognate with Gk. aöpıov), Irish im-
barach, Welsh a-vory (Welsh bore ‘morning’), Lat. cras (whether cognate with Avest. suram
‘arly’ or with Ved. sárvarī ‘dusk’, Hitt. kari ‘early’, kariwariwar ‘at daybreak” ), Goth. du
maürgina (whether root-related with Lith. márgas ‘bright, variegated’ or OCS mrakü ‘dim-
ness, darkness”).

Occasionally a pendant relationship ‘evening’ ` ‘yesterday’ is found, e.g. Lith. väkaras :
väkar, or Russian veter : veerd, or German dialectal jennabend or archaic nächt ‘yesterday’;
cf. ‘eve of’ = ‘day before’. But in the main, except in Baltic and Slavic, the overpowering
association of ‘tomorrow’ with ‘morning’ has been unable to dislodge, iet alone supplant
the ancient terms for ‘yesterday’, not only in archaic but also in heavily innovated forms of
Indo-European ?

Klaus Strunk (JF 73. 1968: 309—310) has suggested that the archaism of words for ‘yester-
day’ ties in with the ancient abilities of the Indo-European verb to denote past time, and
that conversely the later nature of terms for ‘tomorrow’ is related to the gradual devising of
future tense expression in the individual languages. If Proto-Indo-European was incapable
of expressing ‘tomorrow’ lexically as well as grammatically, such a semantic lacuna would
have robbed any word for ‘yesterday’ of some of its antonymity, since its only contrast
term would be ‘today’; it could have expressed the past only as related to the present, not
as also opposed to the future, in a binary rather than ternary semantic structure. It is,
however, improbable that verbal grammar is to blame for this state of affairs. More prob-
ably we are dealing with a shift in conceptual emphasis, from objective side-view of event-
sequences to the subject’s own time-immersion, from a ‘third-person’ to a ‘first-person’
approach. Clearly there must always have been objective means of expressing ‘(on) the next

316

day’, such as Hitt. pard UD. KAM-ti or Skt. (a jparedyus; here days are simply lined up in
calendaric fashion, as when Wulfila translates Luke 7.11 &yevero &v rn eins fit came to pass
the day after’ (€&7> literally ‘in sequence’) by warp in bamma afardaga. But in Corinthians
1.15.32 (gaywpev Kal tioper, adpwr yàp anoðvýokoperv) the first-person speaker is very
subjectively involved, and ‘the next day’ turns into ‘tomorrow’: matjam jah drigkam, unte
du maurgina gaswiltam ‘let us eat and drink, for tomorrow we diel”. Anticipation of the
morning after identified the morrow with the coming day and launched the innovational
term.

By contrast, then, the old term for ‘yesterday’ should have been ‘dispassionate’, with no
hint of a ‘dreadful past’ or emotional involvement with ‘all our yesterdays’; it must have
meant simply ‘the other day’. Its well-known inventory of forms comprises Skt. hyas,
Pahlavi dik, Persian di(g)’, Gk.(é/ydéc, dita, Elean cepdc, Alban. dje, Olr. in-de, Welsh
(yd)doe, Lat. heri (hesternus), Goth. gistradagis, OE geostra, giosterdag, OHG gestaron,
ON (1) g&r or (1) giar. Brugmann (1904: 72) postulated for the first part the demonstrative
pronominal stem *gho- or *ghi-, and Vittore Pisani (1925: 637—643) in a youthful opus
supplied a full reconstruction "gh(e)-dies, variations on which have subsequently entered
the handbooks.® As an alternative to such glimpsing of the etymon of Lat. dies, already
Brugmann (1904: 72, fn. 10) suggested that a suffix -yo- might be involved, denoting ap-
purtenance or adjacency to '(to-)day'; Franz Specht (1944: 201—205) later came up with
a formula *gh-ves containing the equivalent of the primary neuter comparative suffix, and
saw its hypertrophied expression in "£h/yJes-t(e)r- (Goth. gistra-, Lat. hesternus) with the
secondary comparative suffix. Specht was also aware that these suffixes can have a binary
contrastive meaning, as in Ved. navyas- ‘new’ vs. sényas- ‘old’, or in Lat. noster vs. vester.
For *gh- he stuck with the pronominal explanation, trying weakly to explain Gk. xdes
and the Celtic forms via some sort of deictically marked allomorph *gh-t-; the meaning
would thus have been something like ‘this (day) as opposed to today’, i.e. ‘the other day’.

Jan Otrebski (KZ 84. 1970: 85—87) inverted the semantic problem completely by insisting
that, in view of OCS vitera ‘yesterday’ beside veteru ‘evening’, the base meaning had to do
with either ‘shine’ or ‘get dark’, hence connecting either Goth. skeinan ‘shine’, Gk. OK LL
‘shadow’ or a root */s/)ghei-(s}- metathesized ad hoc into *ghesi (Lat. heri).

No further progress was possible without squarely confronting the matter of the initial
consonantism, which is part and parcel of the ‘thorn’ problem. Here the many ‘modern’
treatments starting with Paul Kretschmer’s article (1931/32) have been of uneven value.
Some theoreticians, enamored of their own phonological circularities, have either spawned
monstrosities? or reverted to ground zero.!? Others have chosen to stay mostly reportorial
(J. Gunnarsson, NTS 43. 1971: 52--55). Even those who record the most plausible re-
construct, */dh}gh{y Jes, either leave it unanalyzed (Merlingen 1957: 60) or pronounce it
unanalyzable (J. Schindler, Die Sprache 23. 1977: 32). With this formula the inventory can
be presented as follows:

*dhghyes > Olt. (in-)de.

*dhghiyes > Welsh (yd/doe.

*dhéh(y jes > Gk. xOéc!! (cf. e.g. *dhghom > xduv'*).

*(dh)ghyes > Elean oep/dc), Skt. hyas.

*(dh )ghyes-tro- > Goth. gistra-.!3

#/dh j£hyes > ON {1)gær.

317

*(dh)ghiyes > ON (i) gjar (Gunnarsson, NTS 43. 1971: 54—55).
*/dh Jéh(y Jes > Alban. die,
*(dhjgh(y jesi > Lat. heri.

The pattern is consistently that of the locative case of an s-stem noun *dhghyes-. This noun
may contain the primary ‘comparative’ suffix -yes- (as suggested by Specht) appended to a
root *dhgh- in the weak grade. The word basic to the derivation can have belonged to any
stem class (cf. Gk. rax-Us, *8ax-100- > daroo/0}, or *xùbap, kvôw(o)-). The noun in ques-
tion is in all probability the etymon of English day,'* reconstructible variously as *dhogho-
(Goth. dags, ON dagr, OHG tac), *dhöghes- (OE dag), *dhöghr- (DE dögor ‘day’, ON dégr
‘day or night’), *dhöghn- (ON dógn ‘day and night, 24-hour period").? The commonly
assumed cognate is Ved. dhar-, dhan-, dhas-, Avest. (gen. pl.) asngm ‘day’, where an original-
ly marginally accented paradigm (*dhoghr) : (dh)ghn-és (cf. dsrk : asnds ‘blood’) may have
been rehabilitated in Indo-Iranian as a columnal dher : dhnas, perhaps abetted by the syn-
onymous rhyme-term *ayar-, *ayan- (Avest. ayara, gen. ayan); it can likewise involve a
‘movable dental’ in the manner of Ved. dsru vs. Gk. Saxpv ‘tear’, or Lith. i/gas vs. Ved.
dirgha- ‘long’, or Ved. ama ‘at home’ vs. dám- ‘house’.

An opposition *dhöghos or *dhöghr : *[dh)ghfi)yös (with *[dh]éh[i]vés as suffixless loca-
tive) is quite parallel to Gk. küßos or *küßap : *küduos. The ‘comparative’ suffix involves in
origin not a balanced binary pair of opposites but signals the marked member of the oppo-
sition, even as in Greek eis vs. Grepos (Ionic érepos) reflect IE *sems vs. *smtero- (cf.
Homeric xwaAds 8’ Erepov moda ‘lame in one foot’ [vs. the other]). In the same way */dh/-
éh(i]yos denoted the one of two adjacent days which contrasted with ‘now’, thus ‘the
other day’ (cf. Gk. Qarépa ‘on the morrow’). The meaning ‘yesterday’ (rather than ‘the
next day’) is presumably a conventional one, due in large measure to the pre-emption of
futurity by the ‘morrow’ words. The traces of an alternative meaning ‘the next day’ in Ger-
manic need not be desperately explained away (thus e.g. Brugmann 1917: 10—15; Strunk,
IF 73. 1968: 309-310); thus Wulfila’s (Matthew 6.30) himma daga . . . gistradagis trans-
lating onuepov ... aUpwv speaks of hay that is on the stalk one day and ends up in the oven
the next day, and the Eddic (Hamdismäl 30) nù eda i gær wonders whether death will
come ‘now or the next day’; himma daga ... gistradagis opposes *dhogho- to a doubly
characterized figura etymologica */dh )ghvestro-dhogho-, reminiscent of the manner in
which French aujourd hui twice contains derivatives of the etymon of Lat. diés. The literal
meaning is, however, ‘on the day which is the other day in relation to now’.

Notes

1 Resembling onres, Myc. ssa-we-te (< *kıa-Feres) ‘(in) this year”.

Besides Alban. si-vjet (< *Kyéy-wetesi), Lith. Xi-met '(in) this year’ (cf. M. Huld, KZ 98. 1985: 102--
103). Cf. e.g. Finnish tänä vuonna (X. *vuodna), Estonian tánavu(de) (X. tänä voodena) ‘(in) this
year’, besides Finnish ténd pdivdnd, South Estonian tddmbd ‘to-day’ (all in the essive case).

Parallel to Lat. hodié and OHG hiutu, German heute stand Lat. *hori or *héré (< *hô -ior-) and
OHG hiuru, German heuer (< *hiu järu) ‘in) this year’ (cf. Lat. hérnus, hérnotinus ‘this year’s’,
like German heurig from heuer); the ongoing innovational process is seen in Lat. höc annö > Spanish
ogano ‘this year’. |

Perhaps also with fipi-, NépL ‘early., ajp ‘dimness, mist’, aUpn ‘morning breeze’ (cf. P. Kiparsky,
Language 43. 1967: 624-626).

318

> ‘The Hittite words, however, are better connected with cognates of English ‘grey’ such as ON gryiandi

‘dawn’; cf. Milton’s ‘while the still morn went out with sandals gray’. From Lat. mare ‘(early in the)

morning’ (cf. cräs mäne ‘early tomorrow morning’) are derived the Romance words for ‘tomorrow’.

Cf. Lat. heri > French hier (Italian ieri, Spanish ayer), vs. aujourd ‘hui or demain (*dé mane).

*Avest. Zyó' of many sources is a lively ghostword.

E.g. *£hfi)dfj)es and *ghi-diwes in E. Schwyzer (1939: 631 and 326 respectively).

*g?hyés in E. Benveniste, BSL 38. 1937: 144.

i *éhies in J. Kuryłowicz, BSL 68. 1973: 102.

"The detail of Gk. x9 dá remains unclear; both Skt. a-dyd ‘today’ and mpwitd = mpuonv ‘the day be-

fore yesterday’ have afforded material for comparison, on different chronological levels. Most likely,

however, is *dhghis- as zero grade (samprasärana) suffix variant of *dh£h(y)es-, with suffix as in

«pußsa ‘secretly’ (cf. Specht 1944: 205). xddc is a secondary adjectivization of xoà (cf. x9eoc-

voc).

Metatheses of dental and guttural, postulated in such clusters for Greek and Celtic, seem to have

operated sporadically also in non-contact environments: Gk. óákrvAoc X. *BarkUAoc X. *rkabuAÓc

(cf. J. Puhvel, ZF 81. 1976: 26; Puhvel 1981: 350), or Toch. B kantwo ‘tongue’ « *dnghw-. Cf. the

similar instances of labial and guttural (Gk. axen- X. *spek-; Lith. kümsté ‘fist’ <*pnkWst-).

13 Matched by Lat. hesternus «. *(dh)éh(y)es-tro-no- (cf. Skt. hyástana- and Gk. xdeowds).

M The possibility of such a tie-in was first hesitantly suggested by W. Brandenstein (Glotta 25. 1936:
29) and worked out in more detail in a study by M. Durante (1950: 248); Durante (1950: 247)
brilliantly connected Gk. ydéyyouar ‘produce voice’ with Lith. Zvéngiu ‘neigh’ as reflecting (dh)-
éhweng- .

15 Cf. also the vrddhi-like derivational long grade in Goth. fidur-dögs ‘four days old’, comparable to
Skt. satd-Sdradas ‘of a hundred autumns’.

References

Brugmann, Karl, 1904. Die Demonstrativpronomina der indogermanischen Sprachen (Abhandlungen der
Königl. Sächsischen Gesellschaft der Wissenschaften, Philologisch-historische Klasse 22/6) (Leipzig:
Teubner).

Brugmann, Karl, 1917. “Zu den Wörtern für ‘heute’, ‘gestern’, ‘morgen’ in den indogermanischen Spra-
chen”, Berichte über die Verhandlungen der Königl. Sächsischen Gesellschaft der Wissenschaften,
Philologisch-historische Klasse 69/1 (Leipzig: Teubner).

Durante, M., 1950. “Le spiranti dentali indoeuropee”, RL 1: 243-249.

Kretschmer, Paul, 1931/1932. “X9ww”, Glotta 20: 65—67.

Merlingen, Weriand, 1957. “Idg. ‘Pp’ und Verwandtes”, in: Gedenkschrift für Paul Kretschmer, edited by
H. Kronasser (Wien: Wiener Sprachgesellschaft), 2: 49—61.

Pisani, Vittore, 1925. “Antico indiano ‘hyás’ e ‘çyás’”, Rendiconti della R. Accademia Nazionale dei
Lincei, Classe di scienze morali, storiche e filologiche 6/1: 637-643.

Puhvel, Jaan, 1981. Analecta Indoeuropa (Innsbruck: Institut für Sprachwissenschaft der Universität
Innsbruck). |

Schwyzer, Eduard, 1939. Griechische Grammatik (München: Beck).

Specht, Franz, 1944. “Das indogermanische Wort für ‘gestern”””, KZ 68: 201-205.

Verbi forti e verbi deboli in Germanico

Paolo Ramat (Istituto di Glottologia, Università di Pavia)

1. E’ noto che l’antico sistema dei verbi forti (Vf) germanici (germ.) ha subito una note-
vole riduzione in tutte le lingue della famiglia — talvolta in modo addirittura drammatico,
come nel caso dell’afrikaans, lingua creolizzante nella quale una coniugazione perifrastica
ha sostituito pressocché totalmente le forme preteritali tanto della coniugazione forte
quanto della debole. Per lo scopo di questo scritto possiamo tuttavia lasciare da parte il più
vasto problema delle forme perifrastiche che sostituiscono quelle sintetiche poiché esso
concerne tanto i Vf quanto i verbi deboli (Vd) e non costituisce quindi una discriminante
tra i due tipi verbali.

Forme deboli hanno sostituito per estensione analogica molti antichi preteriti (Pret.) e
participii passati (PPa.): ags. s/ep ^ m.ingl. sJep/sleped(e), slepte — ingl. slept, m.neerl. ge-
wesen > neerl. geweest, ted. wob + webte, boll > bellte, buk + backte, per citare gli
esempi di Jacob Grimm che lamentava la perdita delle forme forti, le quali “schöner und
deutscher singen” (cfr. Augst 1975: 274).

I fenomeni implicati in questo combiamento sono stati ampiamente studiati in dimensione
diacronica dalla linguistica storica e in sincronia soprattutto dalla psicolinguistica (mancano,
invece, osservazioni adeguate dal punto di vista sociolinguistico: vd. fenomeni di creolizza-
zione di lingue germaniche).!

Si è constatato che bambini parlanti inglese passano attraverso quattro fasi nell’apprendi-
mento delle forme preteritali:

(a) produzione di pret. irregolari (irreg.) dall’alta frequenza (go, drink, eat, give, can, may,
have, be, ecc.: in una registrazione di 49 ore di interazione linguistica con bambini tra i 16
e i 24 mesi Slobin ha registrato 292 forme di pret.irreg. contro 99 regolari (reg.): vd. Slobin
1971: 220). Queste forme vengono apprese come lessemi (items) isolati: “each irregular
verb [...] is a case unto itself; no rule covers more than few words and in many cases no
more than one word" (McNeill 1970: 85). Grazie alla loro altra frequenza d'uso queste
forme sono prodotte addirittura prima di quelle regolari (vd. Brown 1973: 220);

(b) acquisizione della regola per la formazione del pret. reg.: /ove > love-d, work > work-
ed;

(c) (iper)generalizzazione della regola (b) con eventuale produzione di forme ipercorrette:
I bringed, I goed e anche I done per I brought, went, did. I cambiamenti analogici mirano
ad introdurre forme più regolari, cioè non marcate, prevedibili, o, per impiegare i termini di
P. Kiparsky, esse mirano ad una semplificazione delle regole (Kiparsky 1971: 174) con
giustappunto l’esempio di brought > bringed).

320

Si osserverà che anche i lapsus linguistici mostrano la stessa tendenza ad eliminare apofonie
(umlaut e ablaut): Esper 1973: 58;

(d) acquisizione di ulteriori forme di preteriti irregolari in una fase successiva dell’appren-
dimento linguistico.

Lo stesso vale anche per il tedesco, dove esiste per il parlante e sopratutto per il bambino
che apprende la lingua una dicotomia fondamentale tra forme regolari e irregolari; queste
ultime, esattamente come in inglese, comprendono Vf, AUX, modali: tutti i Verbi non-
deboli: cfr. Augst (1975: 262). E, come in inglese, gran parte dei verbi irregolari appartiene
al ‘basic vocabulary’ di alta frequenza d'uso.

Se questi dunque sono i fatti (qui riassunti sommariamente), è importante per le osserva-
zioni che seguono che non tutte le forme irregolari mostrano lo stesso grado di irregolarità.
Slobin ha osservato che un paradigma quale bring/brought è ripetuto dal bambino in fase di
apprendimento più regolarmente delle forme preteritali basate esclusivamente sull'alternanza
vocalica (give/gave): "it is as if the final dental consonant is enough of an anchor to the
regular rule that the child can tolerate a vowel in past tense inflection" (Slobin 1971: 220).

Lo stesso si può dire per le forme corrette di preteriti in -d con ablaut: fold, sold; infine
forme come kept risultano ancora più facili in quanto si avvicinano molto a quanto ci si
aspetterebbe secondo gli schemi regolari (*keeped). Si noti che in questo ultimo caso la
forma del preteriti coincide con quella del PPa.: la polimorficità del paradigma basato su
tre forme distinte (originariamente quattro) viene così in ogni caso a ridursi — una ten-
denza ben nota nella grammatica inglese: cfr. to cut, to let (una sola forma), to get, to sit,
to find ecc. (due forme). Qualsiasi siano state le evoluzioni fonetiche alle base di tale ten-
denza, essa costituisce il motivo della sostituzione di spoke, broke a spake, brake regolar-
mente alternanti (ags. spraec, braec) sulla base del PPa. spoken, broken: la sostituzione di
I done a I did (osservata in bambini americani) è dunque lo stesso tipo di regolarizzazione,
meno radicale di */ doed (cfr. le forme effettivamente registrate goed, comed, bringed,
ecc.).

Forme ipercorrette (e ipercaratterizzate) come quelle riportate da Lejosne (1982: 164)
sprongde “he sprang”, kwamste ‘he came” per il dialetto di Westhoek (distretto di Dunker-
que) mostrano in un linguaggio di adulti la stessa tendenza a (iper)generalizzare un para-
digma regolare che abbiamo visto prima alla fase (c) dell’apprendimento linguistico.

Questa tendenza è ovviamente alla base anche dell’estensione del paradigma debole ai danni
del forte osservata in diacronia.? A causa degli accidentali sviluppi fonetici del tipo ora
ricordato alla nota 2, i paradigmi forti divengono sovente del tutto imprevedibili: la tenden-
za a sostituirli con forme deboli è da ricondursi alla tendenza universalmente valida in
morfologia “marcato > non marcato” (cfr. Mayerthaler 1980). La tendenza ad abbandonare
il paradigma forte e la sua incapacità (relativa: vd. più oltre) a produrre nuove forme ebbe
come risultato di sostituire all’antica opposizione fra due paradigmi (forte e debole) egual-
mente vitali e produttivi una nuova opposizione tra regolare e irregolare: questa seconda
serie comprende — come osservavo più sopra — tutti i verbi non-deboli (:*. . . eine Klasse
der starken Verben spátestens seit der Aufklàrung gar nicht mehr gibt. [. . .] Eine Gruppe
‘starke Verben’ gibt es in seiner [= des Sprechers] Kompetenz nicht, sondern nur die Unter-
scheidung regelmäßig (d.h. sprachhistorisch schwach) und unregelmäßig. Unter die letzte-
ren fallen in historischer Hinsicht die starken Verben, die Präterito-präsentia, die Wurzel-
und Suppletivverben”: Augst 1975: 263).

321

E’ tuttavia da notare che sprongde e kwamste conservano ancora la vocale apofonica ori-
ginaria: una soluzione meno radicale delle forme completamente regolarizzate comed,
bringed. Sprongde e kwamste si collocano allo stesso livello di t0/4, sold: ablaut + desin.
debole.

Un problema in parte diverso è rappresentato dalle coppie tipo neerl. zwoer/zweerde ‘I/he
sweared”, dolf/delfde ‘i/he delved’ che hanno forme forti e deboli (ma PPa. forte: gezworen,
gedolven). Dal punto di vista diacronico esse rappresentano evidentemente l'estensione del
paradigma debole, ma la persistenza di preteriti forti mostra ancora il conflitto tra la fase
antica e quella più recente: le forme deboli non si sviluppano da quelle forti per evoluzione
fonetica (salvo forse alcuni casi di Vf in dentale tipo neerl. heten: pret. forte *het deb.
heette, scheiden: pret. forte *scheed deb. scheidde): non c’è passaggio fonetico tra zwoer e
zweerde e la creazione di zweerde avviene ‘ex abrupto’ per analogia. Le relazioni fra i due
tipi dovrebbero esser esaminate per ciascun singolo periodo storico in termini di rispettive
frequenze. Le forme forti di più alto rango di frequenza sono state ovviamente le più resisten-
ti ai processi di neoformazione analogica.

2. Molto meno noto è al contrario il fenomeno inverso, cioè l’estensione di forme forti di
pret. e PPa. a verbi originariamente deboli: vd.p.es. fris. erfde > urf (ted. erbte), merk(e)de
> murk (ted. merkte), neerl. beliéde > beleed, wijzde > wees e sande > zond, schande >
schond, fluit(e)te > floot, blink(e)de > blonk, clink(e)de + klonk, ecc. (infin. belién >
belijden, wijzen, zenden, schenden, fluiten, blinken, klinken). A prescindere da quali siano
stati i singoli sviluppi fonetici,* questi esempi mostrano che il passaggio si è verificato
quando la forma fonica della radice verbale mostra una forte analogia con serie tipiche dei
Vf: cfr. p.es.fris. kerven ‘tagliare’, ora con pret. forte e deb. kurf/kerfde!, neerl. kijken
‘guardare’, krijgen ‘ottenere’ ecc. (secondo la I classe forte), vinden ‘trovare’, zingen ‘can-
tare’ (secondo la III classe), ecc. Come osserva giustamente Mossé per l’inglese, ‘. .. des
timbres [...] étaient sentis comme appartenant spécialement aux verbes forts. Ainsi quand
par hasard un verbe faible avait un de ces timbres, avait-on tendance à en faire un verb
fort" (Mossé 1960: 105). Cfr. in ingl. contemporaneo / wore, tore sul modello dei Vf. 7
swore, I bore (infin. to swear, bear)!

In effetti, esperimenti condotti su anglofoni americani adulti hanno mostrato che esiste
ancora una certa competenza linguistica capace di generare nuove forme forti di verbi
inventati come fo gling o to bing (il che corrisponde alla fase d) dell'apprendimento, vista
più sopra), mentre bambini all'incirca di 6 anni rivelano maggiori difficoltà nel creare forme
'regolarmente' apofoniche (fase c): cfr. Berko (1958: 150—177). Analogamente un esperi-
mento chiedeva a studenti tedeschi di restituire l'apofonia di antichi Vf divenuti deboli nel
medio alto ted., quali schmiegen e schweifen: accanto a un 31% di risposte corrette (schmog
geschmogen, come bog gebogen da biegen o bot, geboten da bieten), 30% delle risposte era
costituito da forme analogiche (schwiff geschwiffen, come griff gegriffen da greifen, biß ge-
bissen da beißen, in luogo di schwief geschweifen) e 39% da risposte erronee principal-
mente per quei verbi che risultano difficilmente associabili a una qualche serie fonetica-
mente simile (‘rhyming series’): caso di schroten ‘macinare’, la cui sola possibile associa-
zione può esser data da stossen; vd. Augst 1975: 269—273.

Risulta da questi esperimenti che esiste dunque ancora una sorta di competenza linguistica
dei Vf come sistema grammaticale, basata sull’analogia rimante con verbi più usuali (vd. i
‘rapports associatifs” di Saussure).

322

Dal punto di vista diacronico è facile concluderne che la creazione di nuovi preteriti e PPa.
forti risale ad un periodo in cui il sistema dei Vf era ancora capace di dar origine a nuove
formazioni, di cui l'esempio piü evidente risulta ancora l'antico prestito germ. *screiban
‘scrivere’, flesso come appartenente alla I classe forte. Si tratta della stessa tendenza verso
neoformazioni analogiche che abbiamo visto a proposito dei pret. deboli di Vf: i parlanti
tendono a serie regolari di forme già presenti nella loro competenza (e dotate normalmente
di alta frequenza). Dal punto di vista qualitativo non c’è differenza tra i due tipi di feno-
meni, cioè tra Vf che passano a deboli e Vd che passano a forti.

La creazione delle nuove forme di pret. forte come urf et sim. può esser stata influenzata
dall esistenza in quasi tutte le lingue germaniche — almeno a un certo punto dello sviluppo
storico — di coppie (per i Vf) del tipo fris. griep/grypte (gripe ‘greifen’), swol/swolde (swolle
‘to swell, schwellen’ e il già citato kurf/kerfde (kerve ‘to carve’), neerl. stiet/stootte (stoten
‘stossen’), zwoer/zweerde (zweren ‘schwören, to swear’, e anche korf/kerfde (kerven ‘to
carve’), dolf/delfde (delven ‘to delve’), krees/krijste (krijsen ‘gridare’), ecc. Queste coppie
possono aver influenzato la creazione di coppie analoghe anche per i Vd dalla struttura
fonologica rimante con modelli forti: cfr. ted. pries, glich (da preisen, gleichen sul modello
di bleichen -- che ha blich/bleichte! —, reissen, schmeissen ecc.), frug/fragte, lud (/ladete)
sul modello di fahren, graben, backen — che ha buk/backte! — ecc.

Questa serie è evidentemente meno ricca di quella opposta (cioè Vf > Vd); comunque essa
esiste in diverse tradizioni linguistiche germaniche (neetl., fris., ted., nynorsk*) e occorre
notare che le forme forti analogiche di solito non sono attestate per i periodi più antichi:
mancano esempi in gotico, antico sass., antico fris., antico alto tedesco; l’ags. conosce
solo il caso di rinan (denominativo da regn ‘pioggia’). Ciò testimonia di una lunga vitalità
del modello forte, o addirittura di una sua ripresa: nel paragrafo seguente cercherò di
fornirne i possibili motivi. Per il momento sarà da concludere che dal punto di vista sincro-
nico e tipologico coppie del tipo murk/merk(e)de debbono esser considerate allo stesso
livello di griep/grypte (vd. schema).

Si prenda anche una forma come Westhoek donste (accanto a danste, dal Vd dansen ‘to
dance’: Lejosne 1982: 163): la si può spiegare solo riferendosi a un momento in cui la III
classe dei Vf (con schema radicale (C)VNC-) era ancora vitale al punto da costituire un
modello per altri verbi. Donste deve esser considerato allo stesso livello di sprongde (Vf che
tende verso una sua regolarizzazione nei deboli con apofonia + desin. debole), per quanto il
suo movimento sia opposto a quello di sprongde; cfr. anche prak (/prekte), joeg (Jjaagde)
da preken ‘to preach’, jagen ‘cacciare’, ted. jagen col pret. dialettale jug) e esloeren (/eslierd
‘scivolato’, PPa. da sliern: vd. Lejosne 1982: 163): in tutti questi esempi siamo di fronte
non a innovazioni lessicali più o meno arbitrarie ma a formazioni analogiche tendenti verso
paradigmi regolari, evidentemente sentiti ancora come produttivi. Molti dialetti tedeschi
conservano l’apofonia dei Vf meglio del tedesco standard (cfr. Schirmunski 1962: IV, 1):
nessuna meraviglia quindi che Lejosne possa citare forme forti come winke, gewunk (Saar-
burg) per winkte, gewinkt o casi di conservazione di forme m. alto ted. quali belle, geboll
(Saaralbe) per bellte, gebellt (:Lejosne 1982: 370sg.).

In conclusione il materiale linguistico attestato è una chiara prova della tendenza a costi-
tuire serie regolari di derivazione. L'estensione analogica poteva verificarsi in entrambe le
direzioni (vd. specialmente la forte attrazione esercitata dalla III classe dei Vf). Coppie del
tipo korf/kerfde e erfde/urf sono indicative della stessa tendenza a regolarizzare i paradigmi

Ge-spann
song
verholen
strewn

LESSICALIZZA- Verwandte

ZIONE

E a

lessemi
isolati
(PPa.)

I was
I went

suppleti-
vismo

I flew
I sang

3 gradi
apofonici

Schema

I done
I spoke

2 gradi
apofonici

sprongde

urf (/erfde) sailed

murk (/merk(e)de) i *bringed
I let dolf (/delfde) ae
I hit korf (/kerfde) dachte *goed

= MORFOLOGIZ-
= ZAZIONE

l grado forte e debole apofonia
apofonico debole e forte +-d/-t

ETE

324

e rappresentano verosimilmente il punto di contatto tra le due serie. Dal punto di vista dia-
cronico, infine, la maggiore regolarità del tipo deb., con un minor numero di cambiamenti
fonetici alteranti la forma, ha fatto sì che questo tipo vincesse sull’altro.

3. E' in realtà diventato impossibile prevedere quale vocale apparirà nelle alternanze di un
Vf in una lingua come l’inglese, a causa delle troppe trasformazioni fonetiche intervenute
rispetto alla forma che ci si attenderebbe dal punto di vista etimologico (cfr. n. 2). E tuttavia
si è visto che paradigmi forti non solo sono presenti in tutte le lingue germaniche moderne
(salvo l’afrikaans), ma che questi sono stati capaci fino ad epoca recente di attrarre ana-
logicamente numerosi Vd. Come si spiega un simile stato di cose?

Nel momento in cui si ricercano le cause di un determinato sviluppo linguistico si è inevita-
bilmente condotti a prendere in esame anche la dimensione psicolinguistica. Ed infatti è
possibile collocare le forme verbali che abbiamo sin qui esaminate lungo una scala i cui
estremi sono rappresentati da ‘grammatica’ e ‘lessico’ e lungo la quale agiscono dinamica-
mente due principii, appunto 'grammaticalizzazione' e ‘lessicalizzazione’ dei quali è evi-
dente la dimensione psicolinguistica.

La tipologia del fenomeno esaminato oscilla quindi per l’insieme dei verbi germanici fra
questi due poli cosicché, per riprendere uno spunto suggestivo di Henry M. Hoenigswald,
“the typology of [...] structural changes is [. . .] constrained by the typology of possible
or probable structures” (Hoenigswald 1963: 45).

Uno dei risultati più interessanti dell’analisi condotta su vasta scala da Lejosne è che emerge
dai varii sistemi dei Vf la tendenza ad una specie di polarizzazione delle vocali: vocali
posteriori e mediane ([u], {o], [a]}) che statisticamente sono le più frequenti nelle forme del
preterito forte (in accordo coi paradigmi delle antiche 7 classi forti), sono sentite come
tipiche del pret. (o passato che sia); vocali anteriori ([e], [i]) come tipiche del presente.’
“Dans ces conditions le phénoméne de polarisation [cioé vocali anteriori per il presente vs.
vocali posteriori per il preterito] apparait comme l'expression d'un refus de l'arbitraire, le
maintien des oppositions fondamentales, la pérennisation d'un systéme bipolaire ou tripo-
laire [se includiamo anche il PPa.] servant de cadre général": Lejosne 1982: 969.$

Su questa base psicolinguistica comprendiamo meglio le neoformazioni come spoke, broke
o come urf, frug, gewunk. E più in generale si può affermare che l’apofonia vocalica è sen-
tita ancora come un segno privilegiato della flessione verbale (a differenza di quella nomi-
nale: vd. sùbito più sotto), sufficiente in sé a differenziare tempi (e modi!) anche in man-
canza di altri morfemi (:/ give” I! gave; ich komme“ ich kam” ich käme).

Ciò risulta particolarmente notevole, poiché l’introflessione è fenomeno sporadico nelle lin-
gue germaniche (il tipo tooth ~ teeth non è più che un relitto in inglese, e anche in tedesco
il tipo Vater ~ Vater, per quanto molto meglio rappresentato, non è più produttivo). In-
oltre si deve tener conto della tendenza generale, propria di tutte le lingue germ., ad abban-
donare la polimorfia dei morfemi: quest’ultima è caratteristica di lingue fortemente flessive
(vd. p.es. il genit. sing. lat. «e, -i, -is, -Us: ‘morfemi sinonimici’ secondo la terminologia di
Skalicka 1966); lo é molto meno per lingue analitiche quali sono diventate quelle germa-
niche. La persistenza dei Vf nel sistema verbale delle lingue germaniche, non solo come
‘tokens memorizzati ma anche come ‘types’ tuttora vitali, contraddice dunque la tendenza
generale (‘drift’) dell evoluzione tipologica delle lingue germaniche e prova ‘e contrario’ la
centralità del sistema dei Vf.

325

L’opposizione comunemente accettata (e peraltro corretta) tra forme lessicalizzate, irrego-
lari e imprevedibili da un lato e forme grammaticali, regolarmente producibili secondo
schemi fissi, dall’altro — opposizione che abbiamo visto ora valida anche a livello psicolin-
guistico — deve essere precisata e sfumata: il sistema dei Vf non è solo un relitto di uno
stato linguistico tramontato, ma è ancora una parte vitale della morfologia e le sue alter-
nanze vocaliche sono sentite ancora come sufficientemente caratterizzanti (iconiche) in
modo tale da proteggere i ‘tokens’ più ricorrenti dalla scomparsa.

Ii polo a sinistra della scala non è occupato allora da forme verbali ‘stricto sensu’; bensì da
forme originariamente collegate al sistema verbale ma i cui legami con questo sono divenuti
via via più deboli. Oltre ai sostantivi verbali (tipo song, Trunk, Ge-spann), queste forme
sono rappresentate specialmente dai PPa. tipo afr. (die) afgestorwenes ‘(i) morti’, (die)
aangesprokenes ‘(gli) interpellati’, ted. (die) erkorenen ‘(gli) eletti’, (die) Verwandten ‘(i)
parenti’, neerl. verholen ‘nascosto’, a. isl. flókenn ‘confuso’, isl. aldinn ‘vecchio’, ingl.
cloven (-footed, hoof) ‘scisso’, shorn ‘privo’, ecc.ecc. Simili forme isolate hanno spesso al
loro fianco un corrispondente PPa.reg.: cfr. neerl. verheeld, ingl. cleft, sheared. Le forme
forti (e quindi irregolari) appartengono di solito alla lingua letteraria, e spesso si tratta di
arcaismi come nel caso di formule stereotipe (vd. p.es. afr. met gebroke hart vs. n gebreekte
been); ovvero il verbo è praticamente scomparso, come nel caso del ted. kiesen. Possiamo
così trovare ancora PPa. forti mentre il pret. ha assunto forma debole: neerl. weven ‘to
weave’, weefde ma ancora geweven; wreken ‘to wreak’, wrekte ma ancora gewroken, schei-
den ‘dividere, scheiden’, scheidde ma ancora gescheeén ‘divorziato’, ecc.

4. Riassumendo: c’é una specie di equilibrio (dinamico, non statico) tra forme regolar-
mente producibili e predicibili che possiamo quindi considerare come generate dai mec-
canismi grammaticali da un lato e forme irregolari non-predicibili che sono immagazzinate
come tali nella memoria dall’altro. Le forme dei Vf (e quelle dei Vd che si comportano
come forti) sono scaglionate lungo l’asse di maggiore o minore grammaticalità. Possiamo
considerare questo equilibrio come il risultato della dialettica fra i due principii opposti del
minimo sforzo (che tende alla regolarità della grammatica) e la simbolicità espressiva (che
tende verso forme non-motivate ma chiaramente riconoscibili, da immagazzinare nella
memoria).”

Ritroviamo pertanto anche all’interno del sistema verbale la ben nota oscillazione tra i due
poli che è alla base dei fenomeni di lessicalizzazione di forme appartenenti alla morfologia
e viceversa di morfologizzazione di lessemi (caso, p.es., del gr. phrilos ‘amico’ — aggettivo e
anche sostantivo, ma originariamente ‘proprio’, aggett. pronominale -- e inversamente del
celt. priatus (< lat. privatus) ‘proprio’ (vd. gall. priawt) ma originariamente ‘sposo’ (vd.
bret. priet ‘sposo’: cfr. Markey 1985).

Fra i due poli, ‘grammatica’ e ‘lessico’, c'è una scala con possibilità di spostamenti (verso

destra e verso sinistra) cosicché è giusto parlare di equilibrio dinamico e, piuttosto che di
‘lessico’ e ‘grammatica’, di processi di lessicalizzazione e grammaticalizzazione.

326

Note
! Una descrizione molto accurata dello ‘status quaestionis’ nella ricerca psicolinguistica è costituita
dalla ‘Thése pour le Doctorat d’Etat’ di Jean-Claude Lejosne (1982). Il presente articolo deve molto
alla lettura di questa ‘Thése’.
Troppi accidenti dell’evoluzione fonetica hanno invero contribuito ad alterare la chiara regolarità del
vecchio sistema apofonico germanico. Si prenda p.es. un verbo quale il fris. fsjen ‘trascinare’ (< germ.
*teuhan-, del got. tiuhan ecc.). Nella forma attuale troviamo: assibilazione di f- nell'affricata ts-
davanti a /-; eu >iu >i, j; -h- >; ma questo -h si mantiene in posizione finale: pret. sing. teach; nel
plur. feagen sono visibili ancora gli effetti dell’ ‘alternanza grammaticale’ (Legge di Verner); ma il
PPa. tein non ne mostra alcuna traccia e l’apofonia non è in e/2}, come normalmente ci attenderem-
mo per un verbo della II classe (cfr. frieze/fear/ferzen ‘to freeze”). Gli esempi di alterazione rispetto
ai paradigmi originarii potrebbero facilmente moltiplicarsi.
Si può pensare che in alcuni casi le forme forti si siano sviluppate sulla base della corrispondente
debole quando questa era foneticamente assai simile: vd. p.es. sande [zando] o l'inserzione di -d- nel
tipo belién. Ma nella gran maggioranza dei casi (vd. il già cit. zwoer/zweerde) non c'é possibilità di
passare foneticamente dall'una all'altra forma.
P.es. jog da jage ‘cacciare, jagen’, che però è un prestito dal basso ted. (cfr. neerl. joeg, ted. dial. jug),
skauy da skyve ‘spingere, schieben’: vd. O. Beito 1970: $ 380sg.
Lejosne (1982: 966) riporta i risultati dell'analisi statistica condotta sul.ted. da E. Mater (1966 —72:
V 108—112):
Vf pres.: e 50, ei 41, i 34, d 5, ie 24:u 1,0 2:

pret.:a 62,0 50, u 14 (vs. 53 forme con vocale anteriore).
Lejosne cita anche le statistiche che mostrano come la quantità d'informazione fornita da vocali
posteriori in inglese sia decisamente più alta rispetto a quella delle vocali anteriori (3,14 vs. 1, 36
bits: vd. Hood Roberts 1965: 40). Risultati analoghi per il ted.: voc. post. 2,63 bits vs. voc. anter.
1, 54 bits (vd. Meier 1964).
Si veda l'idea di Lyons (1974: 90) di considerare la lingua come un sistema omeostatico di auto-
regolamentazione basato sul principio del minimo sforzo, che tende a raggiungere l'efficienza mas-
sima del sistema stesso, e il desiderio di essere compresi chiaramente. Non è di certo una fortuita
coincidenza che la stessa situazione di rilevanti irregolarità nei ‘tokens’ più frequenti la si ritrovi in
altri sottoinsiemi della grammatica (si pensi p.es. ai comparativi tipo better, best, ai pronomi). Per il
ted. vd. Augst 1975: 274, nota 6.

Bibliografia

Augst, G., 1975. Untersuchungen zum Morpheminventar der deutschen Gegenwartssprache (= For-
schungsbericht des IDS, Bd. 25) (Tübingen: Narr).

Beito, O., 1970. Nynorsk Grammatik (Oslo).

Berko, J., 1958. “The Child's Learning of English Morphology", Word 14: 150—177.

Brown, R., 1973. A First Language. The Early Stages (London: Allen & Unwin).

Esper, E.A., 1973. Analogy and Association in Linguistics and Psychology (Athens (Georgia: Univ.
Press).

Hoenigswald, H.M., 1966. Are there Universals of Linguistic Change?, in J.H. Greenberg (ed.), Uni-
versals of Language (Cambridge (Mass.: The M.I.T. Press), 2nd ed. 30--52.

Hood, A. Roberts, 1965. A Statistical Linguistic Analysis of American English (The Hague: Mouton).

Kiparsky, P., 1971. Linguistic Universals and Linguistic Change, in E. Bach — R.T. Harms (eds.), Uni-
versals in Linguistic Theory (New York: Holt) 171—202.

Lejosne, J.-Cl., 1982. Verbes forts et verbes faibles, verbes irréguliers et réguliers. Contribution à l'étude
diachronique et synchronique du système verbal des langues germaniques du Westique. Thése pour le
Doctorat d'Etat, Université de Paris-Sorbonne (Paris IV) (non pubblicato).

Lyons, J., 1974. Introduction to Theoretical Linguistics, 7th ed. (Cambridge: Univ. Press).

Markey, Th.L., 1985. Possession and "Own" (in Indo-European and Beyond), in H. Seiler/G. Brett-
schneider (eds.), Language Invariants and Mental Operations. Intern. Interdisciplinary Conference
held at Gummersbach/Cologne, Sept. 18—23, 1983 (Tübingen: Narr) 124—134.

327

Mater, E., 1966—1972. Deutsche Verben (10 vols.) (Leipzig: V.E.B.).

Mayerthaler, W., 1980. Morphologische Natürlichkeit (Wiesbaden: Athenaion).

McNeill, D., 1970. The Acquisition of Language. The study of developmental psycholinguistics (New
York: Harper & Row).

Meier, H., 1964. Deutsche Sprachstatistik, rist. (Hildesheim: Olms).

Mosse, F., 1960. Manuel de l'Anglais du Moyen-Age: I. Vieil-anglais (Paris: Aubier).

Schirmunski, V., 1962. Deutsche Mundartkunde, trad. ted. (Berlin: Akademie Verlag).

Skalička, V1., 1966. Ein ‘‘typologisches Konstrukt” TLP 2:157--163.

Slobin, D., 1971. On the Learning of Morphological Rules. A reply to Palermo and Eberhart, in D.
Siobin (ed.), The Ontogenesis of Grammar. A theoretical symposium (New York: Academic Press).

Sonderfall Griechisch?

Ernst Risch (Universität Zürich)

1. Bei der Beschäftigung mit der griechischen Sprache und ihrer Geschichte wird man das
Gefühl nicht los, daß hier manches Einmalige vorliegt, etwas, das anders ist, als wir es von
anderen Sprachen her kennen, oder sich wenigstens in einer besonders eindrücklichen Weise
manifestiert. Man ist daher gerne geneigt, von einem Sonderfall zu sprechen. Nun ist jede
Sprache etwas Besonderes und Einmaliges, gewissermaßen ein Sonderfall. Aber wahrschein-
lich ist es die griechische in einem stärkeren Maße als manche andere.

Für die historische Sprachbetrachtung ist es singulär, daß wir hier die Entwicklung von den
mykenischen Tontafein des 14. und 13. Jahrhunderts bis heute, also über eine Zeitspanne
von etwa 3300 oder 3400 Jahren, unmittelbar verfolgen können. Verschiedene andere
Sprachen sind zwar schon früher schriftlich fixiert worden; doch sind sie im Laufe der Zeit
ausgestorben. Die arischen Sprachelemente im churritischen Mitannireich sind, so wertvoll
und interessant sie auch sind, noch keine wirklichen Texte. Im Falle des Chinesischen er-
schwert das völlig andersartige Schriftsystem, sich ein Bild von den Veränderungen der
Sprache zu machen.

Bereits die mykenischen Texte lassen eine Differenzierung des Griechischen in verschiede-
ne Dialekte erkennen, die dann im ersten vorchristlichen Jahrtausend dank zahlreichen Dia-
lektinschriften in einer einmaligen Weise greifbar wird. Freilich hat von diesen altgriechi-
schen Dialekten nur sehr weniges das Ende des Altertums überlebt; von vereinzelten Aus-
nahmen abgesehen, gehen die modernen Dialekte nämlich alle auf die Koine zurück, die
sich seit der hellenistischen Zeit aus dem attischen Dialekt entwickelte. Die neuiranischen
Sprachen und Dialekte setzen mehr alte Dialektunterschiede fort, doch ist das Altiranische
nicht so früh und lange nicht so gut bezeugt.

2. Zweierlei fällt beim Griechischen auf: erstens, daß bereits das Altgriechische, nach dem
Hethitischen (bzw. Anatolischen) die frühestbezeugte indogermanische Sprache, sich durch
verschiedene Merkmale deutlich von den anderen Sprachen abhebt, und zweitens, daß das
Neugriechische diese Merkmale mit einer auffallenden Treue nicht nur bewahrt, sondern
vielfach sogar weiter entwickelt hat.

2.1. Das gilt vor allem für das Verbum. Das Altgriechische ist eine typische Aspektsprache,
indem hier scharf zwischen durativem (oder linearem, präsentischem) und aoristischem
(oder punktuellem) Aspekt unterschieden wird! , wozu als dritter Aspekt das Perfekt (zu-
nächst ausschließlich Zustandsperfekt?) kommt, das freilich eine eher periphere Rolle
spielt. Wir haben allen Grund anzunehmen, daß dieser Aspektcharakter aus der Grundspra-

330

che ererbt ist’. Wenn man nur die Formen betrachtet, ist die Übereinstimmung mit dem
Indoiranischen sehr groß; doch verlieren Aorist und Perfekt im Altindischen schon früh den
Aspektcharakter und werden immer mehr zu Vergangenheitstempora. Ähnliches gilt auch
fürs Iranische. Im Latein fällt der Aorist mit dem alten Perfekt zu einer neuen Kategorie
zusammen, die man ebenfalls Perfekt nennt, die aber in der Regel nicht einen Zustand, son-
dern den Abschluß der Handlung oder deren Vorzeitigkeit ausdrückt. Eine Temporalisie-
rung der Aspektkategorien läßt sich auch sonst beobachten. Im Gegensatz dazu hat das
Griechische bis heute nicht nur den Gegensatz Durativ/Aorist bewahrt^, sondern hat ihn
sogar noch ausgebaut, indem es ihn auch aufs Futur ausgedehnt hat.

Ausgedrückt werden die Aspekte im Neugriechischen wie im Altgriechischen in erster Linie
durch besondere Stammformen, z.B. ypapw “ch schreibe! — Aor. éypaya, 8évco "ich bin-
de’ — é6eoa, uévco» "ich bleibe! — éuewa, uaOatvo» ‘ich lerne’ — éuada, voBoüua "ich fürch-
te mich' — voffjónka? . Wie diese Beispiele zeigen, ist nicht nur das Prinzip als solches be-
wahrt, sondern es sind in vielen Fällen sogar die einzelnen Formen praktisch unverändert
geblieben. Die Stammformen bleiben also eindeutig Aspektstämme, während sie sich in den
andern Sprachen im allgemeinen zu Tempusstámmen entwickeln.

2.2. Dagegen werden die eigentlichen Tempora Gegenwart (Präsens) und Vergangenheit
(Präteritum) im Griechischen durch ganz andere Mittel ausgedrückt, nämlich durch beson-
dere sog. primäre, bzw. sekundäre Personenendungen und zweitens (bei der Vergangenheit)
durch ein zusätzliches Präfix, das sog. Augment®. Beides gilt auch fürs Neugriechische,
doch ist das Augment in der Regel nur dann erhalten, wenn es akzentuiert ist. Dafür ist das
Prinzip der besonderen Endungen auf alle Personen ausgedehnt worden, z.B.

Präs. Sing. Beéevw, SBévers, dévet, Plur. Bévovue, Sévere, Sévovv
Impf. édeva, Ebeves, deve, 6évaue, Gévare, Édevar
Aorist ébeoa, Zëegec, Ed EE, OÉOQUE, Sdé€aare, &6 eoav.

Die Unterscheidung von primären und sekundären Personalendungen ist also im Neugriechi-
schen — im Gegensatz zur Entwicklung in anderen Sprachen — nicht nur beibehalten, son-
dern weiter ausgebaut worden’.

Zu bemerken ist noch, daß die primären Personalendungen wie im Altgriechischen auch für
den Konjunktiv verwendet werden, der beim Durativ mit dem Indikativ Präsens zusammen-
fällt, aber beim Aorist besondere Formen hat. Das eigentliche Moduszeichen ist aber eine
Partikel, meist va (< iva). Diese Form ist unter anderem deshalb wichtig, weil sie auch zur
Umschreibung des fehlenden Infinitivs dient. Mit der Partikel da (< YEAw iva) verbunden
bildet der Konjunktiv das neue Futur, z.B. durativ da ypdypw, aoristisch dà yodyw.

2.3. Als weiteres Erbstück kommt dazu, daß sogar noch im Neugriechischen durch die Per-
sonalendungen auch die Diathesen ausgedrückt werden. Außer dem Aktiv gibt es ein Medio-
passiv, das beim Durativ das alte Medium und beim Aorist das alte Passiv fortsetzt. Die Be-
deutung ist teils passiv, z.B. 6 évoua: ‘ich werde gebunden’ (Aor. deönka), teils medial, z.B.
AoUtouat ‘ich bade (mich selbst)’; es gibt auch zahlreiche Deponentien, z.B. gofovuai ‘ich
fürchte mich’ (s. oben).

2.4. Man beachte also die große Aussagekraft nicht nur der altgriechischen, sondern ganz
besonders auch der neugriechischen Personalendungen, die hierin sehr an die hethitischen
erinnern: Außer der Angabe von Personen und Numerus (also des Subjekts) drücken sie

331

noch Tempus (bzw. Modus) und Diathese aus. Dagegen werden die Aspekte grundsätzlich
durch besondere Stammformen ausgedrückt. Das Neugriechische hat aber nicht nur Präsens
und Aorist, sondern auch das indogermanische Imperfekt nach Form und Inhalt im wesent-
lichen unverändert bewahrt (zum neugriechischen Verbum vgl. Seiler 1952).

Während andere Sprachen den Konjunktiv schon früh aufgegeben haben, es sei denn, daß er
— wie im Latein oder Altpersischen — als Futur eine neue temporale Funktion bekam, ist
dieser Modus im Griechischen bis heute lebendig geblieben. Anderseits ist der Optativ, der
sich sonst viel besser als der Konjunktiv hielt, hier verloren gegangen. Wie das Perfekt
(s. Anm. 4) verschwindet der Optativ weitgehend bereits in der Koine der Kaiserzeit
(Schwyzer-Debrunner 1950: 337£.). Auch der auffälligste Verlust, nämlich der des Infini-
tivs, den das Griechische bekanntlich mit den anderen Balkansprachen teilt, bahnt sich in
deser gleichen Zeit anë .

3.1. Auch das Nomen zeigt heute verschiedene altertümliche Züge. Zwar ist der Dativ auf-
gegeben worden, auch das infolge einer Entwicklung, die sich im späten Altertum abzeich-
net?. Sonst sind alle Kasus des Altgriechischen erhalten, sogar der Vokativ, und ein Para-
digma wie

Sing. Nom. àôepyós ‘Bruder’, Akk. &ôecpyó, Gen. adepyov, Vok. adepye

Plur. Nom. — àbepor, Akk. à6epqoUc, Gen. ad eppuv

stimmt nicht nur mit dem altgriechischen völlig überein, sondern erinnert nicht bloß zufällig
etwa an litauisch draügas ‘Freund’, Akk. draüga, Gen. draügo, Vok. drauge, Plur. draugaT,
Akk. draugüs, Gen. draugü, nur daß im Litauischen drei weitere Kasus, Dativ, Instrumental
und Lokativ dazukommen. Dafür ist das Neutrum sehr gut erhalten, und zwar nicht nur bei
der o-Deklination, sondern auch in Paradigmen wie rò mépos ‘Teil’ (Gen. uépous, Plur.
uépn., uepcov) und dvoua Name’ (Gen. dvouaros, Plur. bvduara, bvouärwr)"®. Die Femini-
na und Maskulina der ersten Deklination sind freilich stärker umgebildet worden und sind
mit denen der dritten Deklination zusammengefallen.

3.2. Eine besondere Altertümlichkeit ist aber, daß der Typus der o-stämmigen Feminina
wenigstens in Resten erhalten ist. Die hierher gehörenden Wörter stammen freilich offenbar
aus der schriftsprachlichen Tradition, wie z.B. r} &£oöoc “Ausgang’, ) eioodos “Eingang, Ein-
tritt’ u.ä., ebenso die restituierten antiken Ortsnamen wie 7 Xtoc. Volkssprachlich sind sie
nach dem Vorbild der übrigen Feminina umgestaltet worden, z.B. n napdevo ‘Jungfrau’,
Akk. rnv napdevo, Gen. rns napdevos, Plur. oi tapdéves, Akk. ric rap9Oévec, Gen. rcov
napdevwv'!. In diesem Punkt ist das Neugriechische offenbar altertümlicher als sogar das
Altindische, das — wie die meisten anderen Sprachen — solche Feminina in die a-Stimme
übergeführt hat, z.B. snusd ‘Schwiegertocher’ statt *snusós, gr. vuóc"? .

4. Zusammenfassend darf man sagen, daß das Neugriechische zu den altertümlichsten unter
den heutigen indogermanischen Sprachen gehört. Zwar ist der Wortschatz bereits im Alt-
griechischen stark umgestaltet; aber auf dem Gebiet der Morphologie übertrifft es in man-
cher Hinsicht alle anderen. Auch bei den Lauten ist manches bemerkenswert, so die Erhal-
tung des auslautenden -s oder die nahezu unveränderte Erhaltung, jedenfalls was die Quali-
tät betrifft, der alten Kurzvokale, die übrigens zum Teil auch die verschiedenen Laryngale
fortsetzen? .

332

5. Es stellt sich die Frage, wie diese und ähnliche Besonderheiten des Griechischen allge-
mein und des Neugriechischen im besonderen zu erklären sind. Daß es sich um zufällige,
von einander unabhängige Einzelerscheinungen handelt, wird man wohl ausschließen dür-
fen. Doch werden vermutlich verschiedene Ursachen zusammengewirkt haben.

5.1. Zunächst ist die Ähnlichkeit zwischen dem Neugriechischen und dem Altgriechischen
zu betrachten, welche, was die grammatische Struktur betrifft, bedeutend größer als etwa
die zwischen dem Latein und den modernen romanischen Sprachen ist. Außer den bereits
besprochenen Erscheinungen kann genannt werden, daß der neugriechische Lautstand im
großen und ganzen jenem entspricht, den das Griechische am Ende des Altertums erreicht
hatte. Erhalten ist vor allem, was morphologisch sehr wichtig ist, das auslautende -s, und
wenn gerade in den vorangehenden Jahrhunderten die Diphthonge und die Langvokale
mannigfaltige Veränderungen erfahren hatten, so ist doch festzuhalten, daß normalerweise
Vokale als solche erhalten bleiben und nicht etwa schwinden, so daß die historische Schreib-
weise auch bei veränderter Aussprache beibehalten werden konnte.

Sicher hat der überragende Einfluß der Kirche und des Klerus, der gerade in der Zeit der
Türkenherrschaft besonders wichtig war, entscheidend dazu beigetragen, daß die in früh-
christlicher Zeit entwickelte Form des Griechischen sich in mancher Hinsicht unverändert
erhalten hat'*. Damit erklärt sich sicher vieles, u.a. die Erhaltung der altertümlichen Para-
digmen beim Neutrum, nicht aber die Schaffung neuer Paradigmen bei den Maskulina und
Feminina der ersten und dritten Deklination. Noch weniger verständlich ist, wieso das
Verbum bei Erhaltung vieler Einzelformen und neben zahlreichen größeren und kleineren
Umgestaltungen ein neues Gesamtsystem von einer imponierenden Geschlossenheit errei-
chen konnte.

5.2. Offenbar setzen sich im Neugriechischen Entwicklungstendenzen fort, deren Wurzeln
bedeutend älter sind und zum Teil bis in die frühgriechische Zeit zurückreichen. Das Myke-
nische zeigt uns, daß die Nominalflexion im zweiten Jahrtausend v.Chr. der grundsprach-
lichen noch merklich näher steht als in klassischer Zeit'” : erhalten war noch als selbständi-
ger Kasus der Instrumental, auch waren verschiedene andere Altertümlichkeiten bewahrt.
Doch war z.B. bei den Nomina agentis auf -rnp bereits eine einzige Ablautstufe verallge-
meinert worden, und die Neutra auf -ua (< *-mn) hatten in den übrigen Formen bereits
den Stamm -war- (bzw. mit einer fürs Mykenische typischen Lautentwicklung -mot-)'®.
Schon in vormykenischer Zeit hatte sich also die Tendenz zu größtmöglicher Klarheit und
Einfachheit wenigstens bei den häufigeren Paradigmen durchgesetzt. Diese Tendenz zeigt
sich auch später, indem etwa die Dative auf -oihi und -āhi (z.B. te-o-i /thehoihi/ ‘den Göt-
tern’, ku-na-ke-ta-i /kunagetahi ‘den Jagern’), also mit lautgesetzlichem Wandel des intervo-
kalischen s zu h, nachmykenisch in Analogie zu den Formen der dritten Deklination zu
-oıcı und -Got (so noch altattisch, z.B. rauiaoı) umgestaltet wurden. Wichtig war vor allem
die Reduktion der Kasus auf fünf, wie wir sie in klassischer Zeit bei allen griechischen Dia-
lekten finden. Auf der anderen Seite zerstörte die Kontraktion aneinander stoßender Voka-
le in der ersten Hälfte des ersten Jahrtausends die bisherige Klarheit und Einheitlichkeit der
Paradigmen, was dann vor allem in der Koine zu mannigfaltigen Umbildungen führte, die
sich zum Teil sogar in den neugriechischen Paradigmen fortsetzen.

Die Entwicklung vollzieht sich also in kleinen und kleinsten Schritten ohne eigentliche

Bruchstellen, sozusagen organisch bei gleichbleibender Tendenz, nämlich Schaffung und
Erhaltung einfacher Paradigmen. Da seit dem Ende des Altertums nur geringfügige lautliche

333

Veränderungen eintraten, bestand nur wenig Anlaß, undeutlich gewordene Paradigmen um-
zugestalten!?.

5.3. Über das mykenische Verbum sind wir leider viel schlechter orientiert als über das No-
men. Doch finden wir auch hier Formen, welche denen der indogermanischen Grundspra-
che näher stehen als denen des klassischen Griechisch, z.B. ki-ti-je-si /ktijensi/ ‘sie bebauen’
= ai. ksiydnti, ohne direkte Fortsetzer im späteren Griechisch. Anderseits ist aber manches
typisch Griechische bereits voll ausgebildet. Man hat schöne Belege für die sog. attische Re-
duplikation, z.B. a-ra-ru-ja, wie hom. apapvia (Part.Perf.fem.) ‘gefügt’ oder ara-ro-mo-te-
me-na /ararmotmena/ "zusammengesetzt (von einem Wagen)’, Part.Perf.Med. zu einem Ver-
bum, das im späteren apuolw -Orrw "zusammenfügen’ weiterlebt. Die größten Folgen hatte
wohl die Umbildung des ererbten s-Aoristes zum griechischen oa-Aorist, z.B. de-ka-sa-to
/deksato/ = (è)Sétaro ‘er bekam’. Denn nun war statt eines Paradigmas mit komplizierten
Konsonantengruppen ein wesentlich einfacheres Paradigma gewonnen, das eindeutig aori-
stisch war? und außerordentlich produktiv werden konnte. Man darf annehmen, daß diese
Umbildung wesentlich dazu beigetragen hat, daß das Griechische den Charakter einer
Aspektsprache behalten und sogar festigen konnte.

Dieser oa-Aorist ist also bereits vormykenisch. Die Formen mit intervokalischem, entgegen
den Lautgesetzen erhaltenem s, z.B. myk. e-ra-se /elase/ ‘er trieb’ (= NAaoe), zeigen, daß sie
offenbar jünger sind als der Wandel s > h!?. Da dieser so gut wie sicher älter als die Einwan-
derung der indogermanischen Vorfahren der späteren Griechen in Griechenland ist, die man
um 2000 v.Chr. datiert (Risch 1979: 273f.), wird man die Schaffung des oa-Aoristes etwa
in die erste Hälfte des zweiten Jahrtausends ansetzen können. Zusammen mit den meist
sehr deutlichen Durativ-(Präsens-)Stämmen erlaubten nun die ebenso deutlichen Aorist-
stämme eine klare Unterscheidung der Aspekte und zwar vor allem durch besondere Stamm-
formen (Aspektstämme). Daher konnte zur Unterscheidung der eigentlichen Tempora Ge-

genwart und Vergangenheit das Prinzip der primären und sekundären Personalendungen
nicht nur bleiben, sondern sogar ausgebaut werden (s. 2.2.). Das Nebeneinander beider
Prinzipien, der Stammformen für die Unterscheidung der Aspekte und der besonderen En-
dungen zum Ausdruck von Gegenwart und Vergangenheit blieb jedenfalls erhalten und da-
mit u.a. auch das alte Imperfekt.

6. Aus dieser notgedrungen sehr summarischen Betrachtung ergibt sich, daß offenbar ver-
schiedene Momente zur eigenartigen Stellung des Griechischen und speziell des Neugriechi-
schen geführt haben. Wichtig ist, daß schon sehr früh ein deutliches Streben nach klaren
Paradigmen faßbar wird, welche im Laufe der Zeit zwar immer wieder umgebaut werden
können, aber in ihrer Grundstruktur nur wenig verändert werden. Dazu kommt, daß sich
das Griechische in lautlicher Hinsicht seit dem Ende des Altertums nur wenig verändert hat,
was wohl — wenigstens teilweise — auf den Einfluß der Kirche und des Klerus zurückzufüh-
ren ist, dem auf alle Fälle die Erhaltung mancher morphologischer Altertümlichkeiten zu
verdanken ist.

334

Anmerkungen

Ich meide die leider sehr verbreiteten Ausdrücke imperfektiv und perfektiv, da sie mißverständlich sind
und immer wieder zu irrigen Vorstellungen führen. Natürlich beruhen sie irgendwie darauf, daß z.B.
russ. ja pisal mit dem lat. Imperfekt scribébam und ja napisäl mit dem Perfekt scripsi wiedergegeben
werden können. Aber mit der Frage, ob die Handlung zum Abschluß kommt oder nicht, hat diese
Aspektunterscheidung an sich nicht viel zu tun. Im Übrigen steht im Latein Perfekt (genauer Prae-
sens perfectum) als Gegensatz zu Präsens (genauer Praesens imperfectum) und nicht, jedenfalls zu-
nächst nicht, zu Imperfekt (genauer Praeteritum imperfectum).

Nach der sehr interessanten Theorie, welche W. Cowgill (1979: 25-39) vorlegt, hätte sich das Zu-
standsperfekt aus der noch älteren Kategorie nominal verbs" entwickelt, also aufs Griechische über-
tragen z.B. Kkekpaye, eigtl. ‘er ist ein Schreier’, was der tatsächlichen Bedeutung sogar sehr nahe
kommt.

Vor allem die Tatsache, daß bestimmte Verbalwurzeln nur als Durativ (z.B. *#7es-) oder nur als
Aorist (z.B. *bhewh>-/bhuh>-), etwa auch als Aorist und Perfekt (z.B. *weid-) verwendet werden,
spricht m.E. eindeutig dafür, daß eine zunächst semantisch bedingte lexikalische Unterscheidung
nachträglich, aber bereits grundsprachlich, als Aspekt grammatikalisiert wurde. Die Tendenz, bei je-
dem Verbum möglichst alle drei Aspekte zu bilden, setzt sich evident erst nach und nach (und sehr
unvollkommen) durch.

Das alte Perfekt ist im Neugriechischen verschwunden, und zwar im wesentlichen bereits im späten
Altertum, s. E. Schwyzer (1939: 779), P. Chantraine (1961: 199ff.). Neu dazugekommen ist aber
ein periphrastisches possessives Perfekt vom romanischen und germanischen Typus habeo ligatum,
nämlich ëxw Seudvo (<dedeudvov) und Exw ödaeı (5 écec eigtl. Inf. Aor.). Doch spielt es im Neugrie-
chischen bei weitem nicht die Rolle wie in den romanischen und germanischen Sprachen.

Wenn nichts anderes vermerkt wird, stütze ich mich fürs Neugriechische auf die offizielle NeoveAAnvırn)
ypauuarık (1976), die auf der Mixp NeveAAnvırn ypapparwi von M. Triandaphyllidis basiert.
Spätere Veränderungen, z.B. bei der Schreibung der Akzente, sind nicht mehr berücksichtigt.
Ursprünglich eine selbständige Partikel, vgl. Risch (1975: 476).

Es sind also im Prinzip die Endungen des sigmatischen Aoristes zunächst auf den starken Aorist und
später auch auf das Imperfekt übertragen worden, s. Anm. 18.

Vgl. etwa Burguiere (1960). Doch ist zu beachten, daß im pontischen Dialekt der Infinitiv, wenn
auch in reduzierter Verwendung, bis heute geblieben ist.

Vgl. Humbert (1930), in allgemeinerem Rahmen Seiler (1958: 41--67). Als Zeugnis für den Zusam-
menfall von Genetiv und Dativ verdienen drei christliche Grafitti aus Kyrene Beachtung: SEG 31.
1578 (a) Kúpte, BohƏngov roð 'Iavovapis, (b) Bońðnaov, Kúpte, roð ġaßßi, (c) (= SEG 9. 188) Kupee,
bBontnoov Avaotaciw rō éxovros tràs nepeorepås (sic!). Man hat den Eindruck, daß hier der kyre-
näische Dialekt nachwirkt, bei dem sich der Genetiv auf -w besonders zäh gehalten hat. Wegen des
lautgesetzlichen Zusammenfalls des Dativs auf -coı mit dem Genetiv lag es nahe, auch sonst zwischen
diesen beiden Kasus nicht zu unterscheiden.

Dem alten Typus dvoua haben sich die Abstrakta auf -ouo angeschlossen, z.B. rò 8 éowo ‘das Bin-
den': 6eo(uaroc, 6ecoiuara, 8eoutárcov usw.

Der Typus h mapdevo fehlt in der NeoeAAnvirn) ypaumarırn (s. Anm. 5), findet sich jedoch bei
Mirambel (1939: 56f.).

Außer dem Griechischen kennt noch das Latein einige Feminina nach der o-Deklination, z.B. humus
‘Erdboden’, alvus ‘Bauch’, Baumnamen wie fagus ‘Buche’ usw., áhnlich auch im Oskischen, Vetter nr.
11 tritbim ekak ... tpsannam deded ‘er hat dieses Gebäude machen lassen’, jedoch im Gegensatz
zum Altgriechischen keine Adjektive auf -us zweier Endungen; auch ist nurus ‘Schwiegertochter’
nach dem Vorbild von socrus ‘Schwiegermutter’ in die u-Deklination übergeführt worden. Das álteste
Latein hatte noch lupus femina 'Wólfin', ebenso agnus femina *weibliches Lamm" und porcus fémina
‘weibliches Ferkel’ (also wie griech. 6 und h Innos), was bereits im Altlatein durch lupa, agna und
porca ersetzt wurde. Man kann sagen, daß das Neugriechische hinsichtlich der Erhaltung der o-stàm-
migen Feminina ungefähr auf der Stufe des Lateins steht.

Z.B. e in derw “ich setze, stelle’ als Reflex von fj, a in marepas ‘Vater’ als Reflex von #), o in
Boravı ‘Kraut’ als Reflex von 715.

Dialekte, welche sich unabhängiger entwickeln konnten, zeigen tatsächlich bedeutend größere Verän-
derungen, z.B. im Tsakonischen kaA& = kaAds und xaAodv, d.h. Schwund des auslautenden c, ferner
vielfach Wandel o > e, vgl. auch rò aépe (oepı) < depos ‘Erntezeit’.

333

5 Wie die Nominalflexion in der Grundsprache genau aussah, können wir hier offen lassen. Doch móch-

te ich bemerken, daß ich die verbreitete Ansicht, daß das Altindische das alte Kasussystem unverän-
dert erhalten habe, nicht teilen kann. Vgl. dazu meine Ausführungen (1980: 259 ff.).

Ausführlicher darüber, sowie über die im folgenden genannten Erscheinungen des Frühgriechischen
Risch (1984: 165—187, besonders 174f., 176, 179ff.).

Der erst im Frühmittelalter erfolgte Wandel von v /ü/ zu /i/ führte z.B. zum Zusammenfall der Para-
digmen gaduc ‘tief’ und Setrc (< bets) ‘rechts’.

Die Personalendungen mit dem Quasi-Themavokal a wurden später auch auf den thematischen (star-
ken) Aorist und nicht viel später auch auf das Imperfekt übertragen, z.B. im NT é6aAav ‘sie warfen'
usw., sogar (als variae lectiones) elxapev -av usw., s. das neugriechische Paradigma unter 2.2.

Dagegen ist das s im dazu gehörenden Futur (ursprünglich Konjunktiv Aorist), wo es seit jeher inter-
vokalisch war, lautgesetzlich geschwunden, z.B. Aaw ‘ich werde treiben’ (mykenisch sind die Parti-
zipien da-ma-o-te und de-me-o-te /-ontes/ zu den Wurzeln *demhz- ‘'zähmen’ und *demh j- ‘bauen’
bezeugt).

16
17

References

Burguiére, Paul, 1960. Histoire de l'infinitif en grec (Paris: Klincksieck).

Chantraine, Pierre, 1961. Morphologie historique du grec (Paris: Klincksieck).

Cowgill, Warren, 1979. “Anatolian hi-Conjugation and Indo-European Perfect: Instalment IP’, in: Hethi-
tisch und Indogermanisch, hgg. von E. Neu und W. Meid (Innsbruck: Institut für vergleichende
Sprachwissenschaft), 25 —39.

Humbert, Jean, 1930. La disparition du datif en grec (du premier au dixiéme siècle) (Paris: Klincksieck).

Mirambel, André, 1939. Précis de grammaire élémentaire du grec moderne (Paris: Les Belles-Lettres).

Risch, Ernst, 1975. “Remarques sur Paccent du grec ancien”, in: Mélanges linguistiques offerts è Émile
Benveniste, éd. F. Bader et al. (Louvain: Editions Peeters), 471—479 (- Kleine Schriften 187 ff.).

Risch, Ernst, 1979. “Les consonnes palatalisées dans le grec du HE millénaire et des premiers siècles du
197 millénaire”, in: Colloquium Mycenaeum, Actes du Sixième Colloque International sur les textes
mycéniens et égéens tenu à Chaumont sur Neuchâtel du 7 au 13 septembre 1975, éd. E. Risch et
H. Mühlestein (Neuchátel: Faculté des Lettres), 267—277 (- Kleine Schriften 549ff.).

Risch, Ernst, 1980. “Betrachtungen zur indogermanischen Nominalflexion”, in: Wege zur Universalien-
forschung, Festschrift Hansjakob Seiler, hgg. von G. Brettschneider und Ch. Lehmann (Tübingen:
Narr), 259—267 (» Kleine Schriften 730ff.).

Risch, Ernst, 1984. “Die Ausbildung des Griechischen im 2. Jahrtausend v. Chr.", in: Studien zur
Ethnogenese, Rheinisch-Westfälische Akademie der Wissenschaften, Abhandlung 72 (Opladen: West-
deutscher Verlag), 165-187.

Schwyzer, I:duard, 1939. Griechische Grammatik 1 (München: Beck).

Schwyzer, Eduard/ Albert Debrunner, 1950. Griechische Grammatik 2 (München: Beck).

Seiler, Hansjakob, 1952. Z aspect et le temps dans le verbe néo-grec (Paris: Les Belles Lettres).

Seiler, Hansjakob, 1958. "Zur Systematik und Entwicklungsgeschichte der griechischen Nominaldeklina-
tion", Glotta 37: 41—67.

Zur avestischen Kompositionslehre: a3.- ‘groß’

Jochem Schindler (Universität Wien)

1. Das jav. Kompositionsanfangsglied (im folgenden KA) aX.- kommt überwiegend in Pos-
sessivkomposita (PK) vor und wird in diesen von Bartholomae, Wb. (ähnlich Duchesne-
Guillemin, Les composés de l'Avesta, 1936, 188f.) durch die Adjektiva ‘groß’, ‘reichlich’,
‘stark’, ‘tüchtig’, viel’ wiedergegeben (z.B. aX bazu- ‘mit tüchtigen, starken Vorderbeinen’).
Außerdem finden Bartholomae, dessen Übersetzungen ich hier zunächst gebe, und Duchesne-
Guillemin 62, 76, 91f., 121, 131 aX- vereinzelt in synthetischen Nomina agentis, z.B.
aX barat- ‘viel, reichlich bringend’, synthet. Nomina actionis, z.B. a$.yesti- ‘reichliches
Opfern’, sowie in adjektivischen Determinativkomposita (DK), z.B. a$.paouruua- ‘der weit-
aus voranstehende, erste’.

Seit Bartholomae, ZDMG 50 (1896) 703°, IF 9 (1898) 282f. und Wb. 229 sieht man in
aS.- für gewöhnlich eine Bildung mit adverbiellem *-s zu idg. *me$- ‘groß’'. Als Bedeutung
dieses selbständig nicht belegten Adverbs setzt Bartholomae ‘sehr, viel’ an. Weiters ver-
gleicht er das nur gav. bezeugte Adv. maš < *meĝ-s etwa ‘sehr’ Y. 32.3, 34.9? und führt als
Parallele für die Ablautdifferenz gr. dya- : uéya an.

Ich versuche im folgenden, die synchronen Gebrauchsweisen von aë.- herauszuarbeiten und
nach Möglichkeit zu erklären. Anhangsweise gebe ich eine neue morphologische Einord-
nung von af...

2. Ich beginne mit den PK. Es ergibt sich zweifelsfrei, daß aš- semantisch nichts anderes
als die Kompositionsform von mazänt- bzw. maz- ‘groß’ (und dem mit diesen in der Bedeu-
tung zusammengefallenen mas-)? darstellt. In einem Fall steht a$.- in direktem Gegensatz zu
kasu- ‘klein’ (ta3.dana- : tkasu.dana-), fiir mehrere Endglieder (KE) sind freie Verbindun-
gen mit mazänt- belegt. Besonders wichtig aber ist, daß mazänt- selber nie als KA vor-
kommt, sich also mit oi. in komplementárer Distribution befindet. Damit erhält aš- einen
wohldefinierten Platz innerhalb der avestischen Kompositionsmorphologie.

2.1. af danu- (wohl *aX.dana-) *mit groben Kórnern' Yt. 8.29: us vó apam aó6auuo | apaiti.
aratà jasünti/ *aX.dànanamca yauuanam | *kasu.dànanamca^ vastranam ‘Die Gerinne eurer
Wasser werden ungehindert herauskommen zum? großkörnigen Getreide und zum klein-
körnigen Gras’°. Das handschriftliche danungmca (und damit ein Stammansatz *dänu-
‘Korn’) ist wegen jav. dano.karSa- ‘Korner schleppend’ und der verwandten Formen mpers.
d’n', ved. dhana-, die av. *dand- erwarten lassen, suspekt und beruht vielleicht auf Einfluß
des GPI. danunam “der Danus’ Yt. 13.38 ter.

338

2.2. aX büzu- ‘mit großen Vorderbeinen’ Yt. 14.12: ya uStré paiti *vaôairi/a$. bäzäuÿ
stuui.kaofo ‘(die Kamelstuten,) die der briinftige Kamelhengst beschützt, der große Vorder-
beine und starke Höcker hat’.

2.3. aXama- ‘von großer Kraft, sehr kräftig’ Yt. 13.107 u. Par.^: kainino kohrpa sriraiià /
aS.amaüä huraodaüä ‘in der Gestalt eines schönen, sehr kräftigen, gutgewachsenen Mäd-
chens’. Vgl. die freie Phrase maziSta ama ‘die größten Kräfte’ Y. 13.3.

2.4. aS.aojah- ‘von großer Stärke, Macht’, im Sav. stets Beiwort von druj-. Y. 9.8 u. Par.:
vò janat aZim dahakam/.../+a$aojagham daéuuim drujim [ayom gaedauuiiö druuantam/
yam *aS.aojastamem drujim|/fraca korantat agro mainiiu$ ‘der den AZi Dahaka erschlug,
... ihn, die sehr müchtige Daeuui Druj, den für die Welt Üblen, den Trughaften, den Agra
Mainiiu als die allermáchtigste Druj hervorbrachte'. Y. 57.15: yo janta *daeuuiià drujo/
ax.aojagho *ahùm.maracò ‘der die sehr michtige, lebenszerstòrende Daèuui Druj erschlug’.
Zu gav. taX.aojá s. 4.1, zu aXaojastara-, aX.aojista- s.5.1. Vgl. noch die freie Phrase mazistam
aojo ‘die größte Kraft’ Yt. 14.12.

2.5. taX.hunara- ‘grobe Fertigkeit, Tüchtigkeit besitzend’ Yt. 10.25: +aS.hunarəm? tanu.
ma@dram / bazuS.aojanham radaestam” ‘(den Midra,) der große Tüchtigkeit besitzt, der den
Mantra in sich hat, den starkarmigen Krieger’. Bartholomae s.v. betrachtet das handschrift-
liche a$ahunara- als legitime Graphie von [aX-hünara-/, doch kann a nur einem yh, nicht
aber einem A vorgeschlagen werden (vgl. ni$.haratar- : niS.ayharsdri-).

2.6. a&manah- etwa ‘von grofem Mut’ Yt. 17.13: aesam uSträyhö bailente/saeni.kaofa
ta$.managhò!® Thre Kamelhengste .. en (baiiente?), die spitzhóckrigen!! , sehr mutigen’
(nach Lommel; Bartholomae ‘mit starker Begierde, Brunst’)!?. Das Beiwort bezieht sich
wohl spezifisch auf die Aggressivität der brünftigen Kamelhengste, wie denn auch im fol-
genden — nach einer leider verdorbenen Verszeile — von ihren Brunftkämpfen die Rede ist
(*paratamna vadairiiauuö “wenn sie brünftig kämpfen’). Vgl. auch die eingehende Schilde-
rung Yt. 14.11—13.

2.7. a$.varacah- ‘von großem Ansehen’, Beiname des Turers Franrasiian und des Kauui
Usan: Yt. 19.57f. frayrase türö a$.varacä, Yt. 5.45 auruuö *aS.varocä‘” kauua usa; Az. 2
bauudhi ... aXvaracá yada kauua usa. Die Form ist F. 8 (= 398 Klingenschmitt!*) zu
varacà verdorben, aber korrekt als KBYR wic'hw übersetzt.

2.8. afxVaranah- ‘groBen Ruhmesglanz habend'?, vom heiligen Wort: maóró spantó yo
a&xVarand (Yt. 12.2, syntaktisch in der Luft hángend; V. 22.2,6; V. 19.16 als Akk.);
maüram spantam | aX xYaranaghom (Y. 2.13 u. Par. in Prosa, aber gleichfalls als Achtsilbler
formuliert). Vgl. die freie Phrase maso . . . xVaranagho Yt. 5.96.

2.9. *as.käma- ‘große Wünsche habend’ V. 20.1 Gl. Der NSg. afkamo, nur in K1 und LA,
glossiert yaoxStiuuatam, PU k’mk’wmnd “Wünsche habend’. Bartholomae s.v. übersetzt un-
nótigerweise ‘mit vielen Wünschen'. Zu elam.-apers. Maï-ka-ma s. Anm. 49.

2.10. a$.mi2da- ‘großen Lohn gewährend’ Y. 55.2: t2 (scil. ga0d) no —buiign humitdà
aS.mizda aso.mizda ‘(die Gädäs) sollen uns guten Lohn, großen Lohn, den Lohn des Aša
gewähren’.

2.11. aS.pairika- ‘von großen Pairikas (d.i. von Erzhexen) begleitet’, Beiname des Pitaona

Yt. 19.41: pitaonamca a3.pairikam. Bartholomae gibt als wórtliche Übersetzung ‘der viele
Pairikas hat’, Lommel hat *hexenreich'. Diese Auffassung, nach der aX.pairika- gleichbedeu-

339

tend mit einem *pouru.pairika- wäre, läßt sich zwar nicht strikt widerlegen, setzt aber für
aX.- eine Uminterpretation von ‘groß’ zu ‘zahlreich’ voraus, deren Annahme sonst nirgends
notwendig ist. Daß mazänt- sonst nicht von daeuuischen Wesen belegt ist, halte ich für Zu-
fall!9.

2.12. aXpacina- ‘groes Kochen habend, wo viel gekocht wird’ Yt. 5.130: yada azam
huuäfritö/masa xSadra niuuanäni/as.pacina stüi.baxadra‘' [fraodat.aspa *xVanat.caxra
xSuuaefaiiat.astra/a’.baouruua niSato.pitu hubaoidsi ‘daß ich wohlbeglückt große Reiche
gewinne, wo viel gekocht und kräftig ausgeteilt wird, wo die Pferde schnauben, die Räder
sausen, wo man die Peitsche schwingt(?)'?, wo es viel Nahrung (s. gleich) gibt, wo Speisen
hingestellt sind, wo es gut duftet'. Falls die KE von aX.pacina und stüi.baxoóra konkrete Be-
deutung hatten, kónnte man die beiden Komposita etwa mit ‘wo es große Kochtöpfe gibt’
(vgl. rigved. pacana- ‘Kochgerät’ 1.162.6), ‘wo es kräftige Portionen gibt’ übersetzen.

Bei der Interpretation von masa x3adra als API. mit der Bed. ‘große Herrschaftsgebiete, Rei-
che’ folge ich Bartholomae u.a. Dagegen faßt Kellens, Les noms-racines de l'Avesta (1974)
77, 356 masa xSadra als ISg. auf (‘afin que je . . . sois vainqueur grâce à un grand pouvoir’),
übersetzt aber die Epitheta nicht. Da man annehmen kann, daß av. mas- wie maz- (7 ved.
mäh-) flektiert hat, stellt masa als NAPI. (:rigved. maha[ni] < idg. *megöh,) entgegen
Kellens kein Problem dar (s. noch Anm. 19).

3. Bei zwei Komposita, aX.baouruua- und aX. vandra-, deren KE als Simplicia nicht belegt
sind, läßt sich die Klassenzugehórigkeit nicht ganz sicher bestimmen.

3.1. aX baouruua- “viel Nahrung habend' Yt. 5.130 (s.o. 2.12) und ähnlich Yt. 17.7: tê
naro xXaóra x3aüente/ax.baouruua nióato.pitu hubaoidi/yahmiia ... ‘Diese Herren gebie-
ten über Reiche (ein Reich?)'?, wo es viel Nahrung gibt usw., wo (in welchem?) . . .'. Das
KE von ai baouruua- gehört zu indoiran. *bharua- 'kauen, verzehren' (rigved. bhárvati).
Bartholomae (implizit Wb. 946 u.) und Duchesne-Guillemin 131 sehen in der Bildung ein
adjektivisches DK zu einem Simplex *baouruua- ‘kauend’, das Kompositum bedeute (‘wo
man viel kaut d.i.) ‘wo es viel zu kauen, zu essen, wo es reichlich (feste) Nahrung gibt’
(Wb. 263). Ich postuliere mit Justi, Handbuch der Zendsprache (1864) 44, 208 schon des-
halb ein PK, weil vergleichbare adj. DK?? bzw. synthet. NAg. mit aš- sonst nicht nachzu-
weisen sind. Dazu kommt das Problem einer Verbindung von xšaðra- mit Nomen agentis
(vgl. Anm. 18).

Gleichfalls als PK kann der altpers. PN Gaubaruva-”' aufgefaßt werden (‘Rind[fleisch] als
Nahrung habend’, Typus ved. uksanna- ‘Stiere als Speise habend’)??. Als genauere Bedeu-
tung des erschlossenen iran. *barua- mag man ‘trockene Nahrung’ (Gegensatz vl. av. pitu-
‘gekochte Speise’) vermuten. Zu mitteliran. Verwandten s. Bailey, Dictionary of Khotan
Saka (1979) s.v. bamrai.

Mit av. as baouruua- ist weiters rigved. subharva- (zweimal von Rindern) zu vergleichen,
für das semantisch ein PK ‘gute Nahrung habend, wohlgenährt’ genauso gut paßt wie ‘gut
kauend, fressend': RV 10.102.5 subharvam ... sahásram /givam ‘ein wohlgenährtes Tau-
send Kühe’ (Geldner); 10.94.3 (wo die Preßsteine mit Stieren verglichen werden): vrksäsya
sakham arunasya bapsatas/té subharva vrsabhah prem aravisuh “Indem sie nach dem Zweige
des rötlichen Baumes schnappen, haben die gutkauenden Stiere (dabei) losgebriillt’ (Ge.).
Statt ‘gut kauend’ kann auch hier ‘gute Nahrung erhaltend’ oder ‘wohlgenährt’ übersetzt
werden (vgl. auch 11d, wo die Preßsteine supivas- feist’ genannt werden). Falls wir es, wie

340

ich glaube, auch bei subharva- mit einem PK zu tun haben, kann seine Verwendung darauf
hinweisen, daß *bharua- wenigstens im Ind. ‘Viehfutter’, speziell vl. ‘Trockenfutter’ bedeu-
tete. Gerade in diesem Fall läßt sich das irreguläre ü in sübharva- leicht erklären, nämlich
als Übertragung aus süvdvasa- ‘gute Weide habend’ (vdvasa- war dann gewissermaßen das
Gegenteil von *bharua-). Problematisch ist natürlich bei söubharva- die Betonung des KA,
die sonst bei PK mit su- sehr selten?3, bei adj. DK indessen häufig ist. Vielleicht hat sich der
Akzent von sübharva- nach einer semantisch verwandten Bildung gerichtet (vgl. z.B. sv-asita-
‘wohlgespeist’ RV 10.28.1), was besonders bei etwaiger ursprünglicher Paroxytonese des
PK verständlich wäre: *su-bhärua- konnte allzuleicht als ‘gut zu verzehren’ mißverstanden
werden.

Ein weiterer Beleg von a$. baouruua- liegt m.E. Yt. 17.8 vor, wo es gleich nach der oben zi-
tierten Stelle 17.7 heißt: aešgm nmand huuidata/gaosuranho histante /a’.paouruua darayo.
upastóe ‘Deren wohlgebaute Häuser stehen da rinderreich, aS.paouruud, zu langer Unter-
kunft’. Im Gegensatz zu Bartholomae (gefolgt von Duchesne-Guillemin 131), der das hand-
schriftliche aS.paouruud halten will (‘die weitaus ersten’), scheint mir B. Geigers Konjektur
zu taf.baouruud (AmoSa Spantas, 1916, 1121), die auch Lommel annimmt, unumgänglich
zu sein. Erstens kann es schwerlich auf Zufall beruhen, daß a3.b’ und aS.p’ innerhalb weni-
ger Zeilen in ganz ähnlichem Kontext aufeinanderfolgen (daß baouruud einmal bewahrt,
das zweitemal aber durch das geläufige paouruuá ersetzt wurde, stellt bei der Überliefe-
rungslage von Yt. 17 kein Problem dar); zweitens findet a$.paouruua- entgegen Duchesne-
Guillemin a.0. in hupauruua- Y. 52.3 keine Parallele: var’hiica add van”hisca afaiiò /
hupauruud vahehi$ apará “die guten Darbringungen und die guten Belohnungen, die guten
früheren, die (noch) besseren künftigen’. Während hupauruua- im KA funktionell ein Ad-
jektiv enthält (‘der gute frühere, der früher gut war'?^), kónnte aX.- natürlich nur die Bedeu-
tung eines Adverbs haben?5.

3.2. a&vandra- *groben Lobpreis habend, sehr gepriesen’ Yt. 19.9 u. Par.: uyram kauuaem
xVaranò jmazdaSatam yazamaide/as.vandram uparo.kairim “Den mächtigen kauuischen
Ruhmesglanz, den mazdägeschaffenen, verehren wir, den sehr gepriesenen, überlegen wir-
kenden’. Bartholomae setzt *aX.vandara- an (PK mit einem *vandara- m.), doch ist erstens
aS.vandra- besser bezeugt? (dazu kommt noch das aus unserem Wort entstellte vandram
F. 8 = 397 Kli.), und zweitens steht vor uparö.kairiia- immer ein dreisilbiges Vorderstück,
vgl. Y. 9.10 aat aniio uparö.kairiiö, Yt. 8.4 barazantam uparo.kairim, Yt. 15.12 vaiiu yo
uparö.kairiiö, das Wort ist demnach fünfsilbig?”. Das vorauszusetzende *vandra- ‘Lob-
preis’ von vand- ‘preisen’ ist m.E. aus einem Adj. *vandra- *preisend'?5 substantiviert, vgl.
tigved. vandaru- ‘Lobpreis’ (4.43.1, 6.4.2), das durch Ellipse aus váco vandáru ‘preisendes
Wort’ (5.1.12) von vandaru- ‘preisend’ (m. 1.147.2) hervorgegangen ist.

Als Alternative zur hier gegebenen Interpretation von as.vandra- als PK käme die als adj.

DK ‘sehr gepriesen’ zu einem *vandra- ‘gepriesen’ in Betracht (so Duchesne-Guillemin
LES,

4. Fürs Gav. haben Insler und Kellens drei PK mit a$.- durch Konjekturen herzustellen ver-
sucht.

4.1. Einen gav. Beleg von aX.aojah- (s.o. 2.4) finden sowohl Insler (Gäthäs 224) als auch
Kellens (Stlr 4, 152ff.) in *af.aojá naidiiüghom Y. 34.8 ‘der sehr mächtige?” den schwäche-
ren’, wo bisher as aojä gelesen wurde. Zur reichlich unklaren Gesamtinterpretation der

341

Strophe verweise ich auf die Diskussion der beiden Gelehrten. Bei as müßte es sich um ei-
nen relativ alten Überlieferungsfehler handeln, da dieses der PU ( YT" 'ist") zugrunde liegt
Manuskripten des [ndischen Videvdad Sada und des Yasna Sada). Wahrscheinlich aus unse-
rer Stelle, aber interessanterweise ohne af, ist Y. 57.10 yada aoja*' ndidiianham ‘wie der
mächtige den schwächeren’ übernommen.

4.2. ta¥.xratu- ‘von grofer Geisteskraft’ Y. 31.9.: 083 @ gau¥ tata ‘aë.xratuÿ/mainiiuÿ
‘Thine was the fashioner of the cow, the spirit of great determination’ (Insler, Gathas 184).
Zur gleichen Auffassung kommt wiederum Kellens Stlr 4.155. Auch hier wurde bisher mit
der PU ( YT’) as gelesen*.

Zu jav. aS.xradßastoma- s.u. 5.2. Vgl. noch die freien Phrasenxratum. . . masitom mazantam
Y. 62.4 und masiid . . . xratuS Yt. 10.107.

A3. Toi änt 'grofe Macht habend' Y. 44.9: Ofauugs *axX.iXti$ mazda ‘someone of great
power like Thee, wise lord' (Insler, Lg 47, 1971, 575’ und Gathas 69, 246; zustimmend
Kellens, StIr 4.152'*). Bisher las man asi3tiX (nach Humbach ‘Anweiser”), was wohl zumin-
dest schon die Lesart der Yasna-Stammhandschrift war (aÿ° nur aÿisti S1 J4 und aÿisriÿ
Je Jm1; S2). Die PU tyc ‘schnell’ beruht auf Verwechslung mit asiXta- ‘der schnellste" Y.
30.10, 34.4 (an beiden Stellen a$° als v.1.).

S. Die mit aX.- gebildeten Superlative erfordern eine eigene Behandlung. In einem oder
mehreren Superlativen zu PK mit a$.- scheint dieses nämlich zu einem Augmentativpräfix
in der Bedeutung des deutschen 'aller- uminterpretiert worden zu sein. In Neubildungen
konnte dann aš- einfach einem schon existierenden Superlativ vorgeschlagen werden. Im
einzelnen bleiben manche Unsicherheiten bestehen. — Der einzige Komparativ wird hier
mitbehandelt.

Bartholomae s. vv. legt den Superlativen jeweils bereits mit aX.- gebildete Positive zugrunde.
Duchesne-Guillemin, der Comp. 131 die Superlative kommentarlos unter den adjektivi-
schen DK einordnet, müfste wohl von Syntagmen à la homer. uéy' dpioroc ausgehen.

5.1. aS.aojastara- ‘noch mächtiger’ V. 9.48: aea drux$ ya nasus aS.aojastara var26aiiete
yada para ahmat as ‘Diese Druj Nasu wächst zu einer noch mächtigeren heran als sie zuvor
(schon) war’. Die Form ist regulärer Komparativ zum PK aš.aojah-. Ob auch im Komparativ
aš.- zu einem Präfix uminterpretiert wurde (etwa im Sinne von deutsch ‘noch’), läßt sich
mangels weiterer Beispiele nicht entscheiden.

Der zugehörige Superlativ ist aš.aojastəma- ‘der allermächtigste’ Y. 9.8 u. Par. (Text oben
2.4.). Wegen der Existenz des Adjektivs aojah- und der zumindest potentiellen des zuge-
hörigen Superlativs *aojastəma- (vgl. ved. tavas-, tavastama-) scheint es nicht unwahrschein-
lich, daß die Neuauffassung von a3.- als verstärkendes Präfix gerade in aS.aojastama- erfolg-
te. Dabei konnte man a3.aojastama- als verstärktes aojastama-, aber auch als eine direkt auf
aojah- beruhende Bildung interpretieren.

Auf neue Weise gebildet ist das gleichbedeutende aX.aojista- im späten Fr.W. 8.2: ha drux$
*a3.aojista ayhat ‘Diese Druj wird die allermächtigste sein’. Hier wurde aë.- dem Superlativ
aojiXta- *der máchtigste' práfigiert.

5.2. aX.xraü(astoma- *der allerweiseste' Yt. 10.141: aojiXtangm asti *aojiXtó / tancistangm
asti *tancisto?? | ba'yanam asti a&xraüfastomoó "Unter den sehr starken ist er der stärkste,

342

unter den sehr kühnen ist er der kühnste, unter den Göttern ist er der allerweiseste’. Als re-
gulärer Superlativ von *aS.xratu- (vgl. 4.2.) ware *aS.xratutama-, vom Simplex xratumat-
*xratumastoma- zu erwarten. Belegt sind stattdessen aXY.xraO(astoma- und xraðßišta- (Yt.
13.80 u. Par., Y. 1.1). Die beiden Formen sind zweifellos direkt dem Paar aX.aojastoma- :
aojiXta- nachgebildet. Einen Positiv *a$.xradßant- (Bartholomae, Wb. 260, 537) hat es nie
gegeben.

5.3. ta$. varadrajastoma- *der allersiegreichste" Y. 9.15: yo aojistó yo tancisto | yo 0faxNiSto
yo asiXto/yO *eX.voroórajastamo?* [abauuat. *maniiuuá. *dàma ‘(Zaradustra,) der der
stárkste, der der kühnste, der der eifrigste, der der schnellste, der der allersiegreichste der
Schöpfung der beiden Geister wurde’. Die gleiche Folge, nur mit litaneihafter Auflösung,
Yt. 10.98 = 135: yo aojisto yazatanam [yo tanciStö yazatanam yo BfaxSiSto yazatanam |
yo asiîto yazatanam/yò +taS.vara8rajastamo* yazatanam /fraxStaite paiti diia zomà
'(Mi?ra,) der . . . als allersiegreichster unter den Göttern auf dieser Erde vordringt’. Zur phi-
lologischen Begründung der Konjektur *aX* s. Seiler, Relativsatz, Attribut und Apposition
(1960) 95f., Humbach, FS Lentz (1974) 91, Kellens, StIr 4.151f. Die Alternative, daß as
zu halten ist und aus einer Stelle stammt, an der die Kopula berechtigt war (so Bartholo-
mae, Wb. 279, Fn. 36), ist weniger wahrscheinlich. Vgl. noch Gershevitch, Hymn to Mithra
248f. In ta$.varadrajastama-, wenn richtig hergestellt, liegt der durch aX.- verstárkte Super-
lativ varadrajastama- ‘der siegreichste’ (von varadrajan-) vor.

5.4. aSadßö.zgatoma- ‘der allerregsamste(?)’ Y. 13.2: ao Ößö.zgatama gauuastriiauuarSto ma
naram aXaonam vastriiehe fSuiianto ratum amruiié ‘Die allereifrigsten, die landwirtschaftlich
tätigsten unter den aSagläubigen Männern rufe ich auf als Ratav des viehzüchtenden Acker-
bauers’ (Wolff). Die Bildung gehört wohl zum Hapax Ofazjaiti Yt. 19.58, nach Gershevitch,
TPhS 1964, 16f. ‘is aroused, startled, upset’. Über die Derivationsgeschichte läßt sich nichts
mehr ermitteln.

5.5. Im selben Stück steht das völlig unklare taf.xra.xVanutoma-. Y. 13.3 (pseudogav.)= Vr.
3.5 (normaljav.): amafasca sponte saoSiiantascà dahifta *arXvacastoma aifiüamatomg
taS.xra.xVanutama®® ‘die Amo$a Spontas und die Saotiiants, die kundigsten, die am richtig-
sten formulieren, die übermächtigsten, die ...sten'. Nach Bartholomae s.v. ist die Form
Superlativ von einem PK *aX.xraxVanu-. Das zugrunde liegende *xraxYanu- m. gehöre zum
Verbum xrághaiia- *erschüttern" und sei eine Bildung wie ved. vagvanu- ‘'Getön’. Als Bedeu-
tung setzt Bartholomae 'der am allermeisten antreibende, anregende’ an.

Das genannte Verbum wáre nur Yt. 10.36 belegt: fra maiSiianam xranhaiiete /spabahe
xruuisiiantahe 'Erschüttert wird die Mitte des blutdürstigen Heeres’ (Wolff), ‘is quaking’
(Gershevitch). Humbach, FS Lentz 87’ vermutet, xrayhaiia- “could belong to the root
kar$ ‘to plough’ just as Orághaiia- to the root tars- ‘to tremble’” (was unter anderem
Schwebeablaut voraussetzte). Er übersetzt ‘he cuts asunder the center of the blood-thirsty
army’. Indessen ist, wie auch Kellens, Verbe 148%, 262, geschen hat, *óránhaiiete zu lesen
(Jarayhiieti H4, das einen von der F1-Linie unabhängigen Text bietet) und zu übersetzen
“In Schrecken versetzt er usw.”.

Dazu kommt, daß das von Bartholomae als morphologische Parallele angeführte ved.
vagvanu- (RV 9.3.5 avis krnoti vagvanum ‘er bringt Getön hervor’) selber isoliert ist und
dem Verdacht unterliegt, eine aus vagnu- ds. unter dem Einfluß von vagvand- ‘geschwätzig’
und einer der Formationen auf -anı- (vgl. 8.21.14 yada krnosi nadanum ‘wenn du das
Schlachtgeschrei erhebst’ [Ge.]) umgestaltete Augenblicksbildung zu sein, s. Renou, EVP
8.51.

343

Die PÜ von taS.xra.xVanutoma-, KBYR hlt'-krt Itwm ‘am meisten Verstand verschaffend’
beweist, daß die überlieferte Wortform so schon dem Übersetzer vorgelegen hat: krt Irwm
will das an kwn- ‘machen’ anklingende x’anutama wiedergeben. An eine (auch in den Vr.
verschleppte) Korruptel — etwa aus *aX.xraófastomg — darf daher kaum gedacht werden.

5.6. a8.xYaratama- ‘der allergefräßigste’ V. 3.20 2 9.49: aX xVarotamaeibiio spanto.mani-
iauuanam dämanam tkərəfš.x”aram kərəfš (statt kohrpam) paiti.nisrinuiiat vaiiam kahr-
kasam ‘Den allergefräßigsten unter den fleischfressenden Geschópfen des heiligen Geistes
überantworte man die Leiche, den Geiervögeln’. Am einfachsten wäre es, in ay. xVaratama-
den Superlativ zu einem synthet. NAg. Sot var. ‘sehr fressend’ zu sehen. Da jedoch genau
vergleichbare Bildungen fehlen (zu aS. barat- s. gleich), muß erwogen werden, ob die Form
nicht vielmehr als mit aX.- verstárktes *xVairiXta- 'am meisten fressend' intendiert ist, wobei
-iXta- durch das in Komposita normale -t2rma- (es würde z.B. *korofs.xVaratama- heifjen) er-
setzt worden wäre.

6. Synthetische Komposita. Wir finden (von dem soeben behandelten a3.x’aratama- abge-
sehen) nur noch ein Nomen agentis und drei eng zusammengehörige Nomina actionis.

6.1. ta¥.barat- ‘Y Yt. 13.23: và taf.barato vd uyraroto/yà huuaratò ya vazarato/ya
taxmárató yà zaoiidrato [yd zaoiid vaghuó(aeSu ‘(les frauuaëis,) qui apportent beaucoup,
qui s'élancent puissantes, qui s'élancent d'elles-mémes, qui s'élancent avec puissance, qui
s'élancent hardies, qui s'élancent dignes d'étre invoquées, qu'on doit invoquer dans les
dangers' (Kellens, Noms-racines 128).

Die traditionelle Übersetzung 'reichlich bringend' ist keineswegs sicher. Auf eine andere
Móglichkeit führt die Interpretation der übrigen Komposita unserer Stelle, die bei zaoi-
igrat- einzusetzen hat. Falls zaoiiarat- nicht ein sonst unbelegtes *zaoiid- ‘Anrufung’ ent-
hält (‘auf den Ruf sich aufmachend’ Bartholomae s.v.), sondern, wie wahrscheinlich, das
auch gleich im folgenden dreimal erscheinende Gerundiv zaoüa-, kann dieses funktionell
nicht Adverb sein: ‘die sich auf anzurufende Weise in Bewegung setzen’ wäre nicht sinnvoll.
Vielmehr muß zaoiia- in diesem Fall die Funktion eines Prädikatsadjektivs haben ‘die sich
als anzurufende in Bewegung setzen ??. Es wäre dann weiter wahrscheinlich — und Kellens
übersetzt entsprechend —, daß die parallelen uyrärot- und taxmärat-” als ‘die sich als starke,
kühne in Bewegung setzen’ zu interpretieren sind (‘Kräftig, kühn sich aufmachend’ Bartho-
lomae s.vv., vgl. Duchesne-Guillemin 80); wyra- und taxma- sind ja geláufige Epitheta der
Frauuasis. Zur vorauszusetzenden Konstruktion *ya uyrä (etc.) iiaronti vgl. z.B. Yt. 5.3
yd amauuaiti fratacaiti ‘die als mächtige vorwärtsfließt”*. Damit eröffnet sich die Möglich-
keit, auch im KA von aS.barat- funktionell das Adjektiv ‘groß’ zu sehen und vorläufig zu
übersetzen ‘die als große bringen’. Man vergleiche dazu besonders Yt. 13.64: yå masiiehiš
ahmat/yà aojiiehi$ ahmat/yà tafiichi$ ahmaāt/. . ./yaða vaca +framruudire ‘(die Frauuašis,)
die größer, stärker, kühner sind ... als sie durch das Wort beschrieben werden können’.
Hier geht den Komparativen von wyra- und taxma- der von mas- = mazänt- voraus, geradeso
wie an unserer Stelle a3.-, die Kompositionsform von mazänt-, vor uyra- und weiterhin
taxma- steht.

Die Interpretation des KE -barat- als “bringend’ ist ebenfalls nicht unproblematisch. Außer
in der Bedeutung ‘reiten’ (die Frauuagis sind aber nicht beritten) kommt verbales bar- nie
ohne direktes Objekt vor, und auch die Vorderglieder sämtlicher jav. synthet. NAg. auf
-barat-, -bara-, -barana- fungieren als Objekte von bar-*. Außerdem ist Unterstützung das

344

einzige, was die Frauuaëis bringen (baron upastam Yt. 13.1 = 19). Nicht ganz mit Unrecht
sagt Herzfeld, Zoroaster and his world (1947) 499 “‘much-bringing’ is nothing”. Auch
der asymmetrische Bau der Strophe — ein Kompositum mit -bərət- gefolgt von fünf mit
-ərət-—ist auffällig. Vgl. dagegen z.B. die mit 6 fraxSti.da yo azuiti.da beginnende Reihe von
acht Komposita mit ‘dd Yt. 10.65. Nimmt man dazu, daß gerade a$.- dasjenige Vorderglied
der Serie ist, nach dem -arat- nicht durch Kompositionssandhi verändert worden wäre,
muß mit der Möglichkeit gerechnet werden, daß die Form des Textverfassers *aZ-arato ‘die
sich als große in Bewegung setzen’ lautete.

6.2. ad.yesti- ‘große Verehrung’ Y. 68.9: aca nö jamiüa auuanhe/as.vestica huilestica/
hufrabaratica zaodranam “Und mögest du uns zu Hilfe kommen aufgrund (unserer) großen
Verehrung und guten Verehrung und der guten Darbringung der Opferspenden”.

as.fraiiasti- ‘große Verehrung’ und aS.frabaraiti- “große Darbringung’ erscheinen in dem
deutlich aus dem eben zitierten Stück umformulierten und mit mehr Parallelismus versehe-
nen Passus Yt. 10.77: dca nö jamüat auuanhe/a3.fräiiastica zaodranam hufraiiastica/
Ya. fraborotica zaodrangm *hufrabarstica. Sie haben kaum sprachgeschichtlichen Eigen-
wert.

Wahrscheinlich ist das isolierte a$,yesti- zu huie$ti-, das einem ererbten und produktiven
Typ angehört, nach dem Vorbild der PK (vgl. humi2dä af.mizda Y. 55.2) aufgrund der
möglichen Auffassung der komponierten zi-Abstrakta als DK hinzugebildet worden. Seiner
Bedeutung nach ist huüesti- die komponierte Version von vayhu- yasna-, vgl. vaghüs
yasnasca (vahmasca) Y. 23.3 u. Par., ebenso wie aš.yešti- die von mazänt- yasna-, vgl.
mazistam yasnam Yt. 1.24.

7. Unsere Materialuntersuchung führt zu folgendem Ergebnis: a3.- ist Kompositionsform
von mazänt- ‘groß’, hauptsächlich in PK (2—4). In Superlativen von PK wie a$.aojastoma-
wurde a3.- bis zu einem gewissen Grad zu einem verstärkenden Präfix (deutsch ‘aller-’) um-
interpretiert (5). Eigentliche adjektivische DK, in denen fiir a5.- die Bedeutung eines Ad-
verbs ‘sehr’ zu erwarten wäre, lassen sich nicht nachweisen. Die dafür angeführten Beispiele
wie aS.vandra- (3.2.) erlauben jeweils andere Interpretationen. Adverbielle Bedeutung wäre
im Normalfall auch bei synthetischen NAg. zu erwarten (vgl. RV 8.46.18 mahi-sväni- ‘sehr
brausend’). Von den beiden einschlägigen Bildungen ist nun die eine, ax xVarotora-, als
Superlativ nicht voll beweiskräftig (5.6.), und die andere, aX. berot-, dürfte funktionell ein
Prádikatsadjektiv enthalten (6.1.) Auch die Abstrakta wie aš.yešti- ‘groe Verehrung’,
die analogisch zu solchen mit hu- geschaffen wurden, können adjektivisch aufgefaßtes a3.-
enthalten (6.2.).

Im Indoiranischen ist der Gebrauch der Kompositionsform eines Adjektivs^" auch in der
Funktion des zugehörigen Adverbs in synthet. NAg. regelrecht (für die anderen einschlägi-
gen Kompositionsklassen gibt das Material zu wenig her). Das gilt auch für solche Fälle,
in denen eine größere formale Differenz zwischen Adjektiv und KA besteht. Vgl. für
Caland-Adjektiva rigved. zuvi- in synthet. NAg. wie fuvi-badha- ‘stark bedrángend' (: Adj.
taväs-), jav. barazüi-aogat “mit hoher (lauter) Stimme sprechend’ (: Adj. baraz-, barazat.,
rigved. Adverb brhät ‘hoch, laut’), xruui-yni- Name einer Daeuui, wenn eigentlich ‘grausam
tótend' (: Adj. xrura-). Freilich ist es bei der spärlichen Verwendung von Adjektivadverbien
im Indoiran. oft nicht sicher, ob das KA der Funktion nach ein Adverb oder ein Prädikats-
adjektiv repräsentiert (letzteres offenbar in xSuuißi.vaza- Yt. 8.37: xuuifi.vazom/yo

345

auuauuat xSuuaéBo vazaite ‘den schnellfahrenden, der so schnell fährt’). Aus historischer
Sicht ist die Regel jedenfalls sinnvoll, da ja die synthet. NAg. mit verbalem KE als idg. Ty-
pus letztlich mit großer Wahrscheinlichkeit auf uminterpretierte PK mit NAct. als KE zu-
rückgehen.

Daß in unserem Fall weder das KA a?.- in eindeutig adverbieller Funktion noch (im Jav.)
ein zu mazänt- gehöriges Adverb belegt sind, könnte nun ganz einfach Zufall sein: Im RV
kommt zwar das Adverb mahi öfters vor, aber außer dem schon genannten mahi-sväni-
gibt es nur noch ein synthet. Adj., mahi-vrdh- ‘großgewachsen’ (7.31.10; vgl. pérvatas
cin mahi vrddho ‘der hochgewachsene Berg’ 5.60.3).

Es besteht aber auch die Möglichkeit, daß es im Jav. ein Adverb ‘sehr, valde’ gar nicht ge-
geben hat. Auch vaghu- ‘gut’ bildet bekanntlich kein Adverb, ‘bene’ läßt sich nur durch
Umschreibung mit figura etymologica ausdrücken (Typus huberotom bar- ‘jemanden gut
halten, pflegen’). In diesem Fall wäre auch das Fehlen von Komposita, die ein Adverb
‘sehr’ der Funktion nach enthalten, verständlich“. Man beachte noch die figura etymologica
(statt eines Adverbs) für ‘maxime’ in kô paoirīm imam zam maziSta x$naoma xSnauuaüeiti
(V. 3.12) Wer befriedigt erstens diese Erde mit größter Befriedigung (d.h. am meisten)?
(ein zweites Beispiel V. 18.61).

8. Wir wenden uns nun kurz der morphologischen Bestimmung von aë.- zu. Bartholomaes
Interpretation, nach der a$.- eine Bildung mit 'adverbiellem" *-s wáre?", ist nicht haltbar.
Wahrend Nominalstimme, die als Kompositionsformen von Adjektiven verwendet werden,
wie oben ausgeführt, auch adverbielle Funktion haben kónnen, trifft das umgekehrte nicht
zu: Adverbia sind nur dann als KA von PK zulässig, wenn sie prädikativ, d.h. in Verbal-
phrasen mit der Kopula verwendet werden kónnen^??, was gerade bei Adjektivadverbien
nicht der Fall ist (vgl. Hoffmann, Aufsátze 2.339ff.). Nicht mit der Kopula verbindbare
Adverbia erscheinen, ihrer syntaktischen Rolle gemäß, in synthet. NAg., NAct. und passi-
ven Verbaladjektiven, sowie allenfalls in adjektivischen DK. Sie können natürlich auch in
solchen formal wie PK aussehenden Bildungen auftreten, deren KE sekundär wie die von
synthet. Komposita verwendbar geworden waren^?. Doch übernehmen sie nicht die Rolle
der ihnen allenfalls entsprechenden Adjektiva in eigentlichen PK.

Wenn somit ein Adverb nicht in Frage kommt, kann *mgs-, die Vorform von a3.-, nur die
schwundstufige Stammform eines s-Stamms *meg-es- darstellen. Die Reduktion, zu der
man formal *mns-dheh,- von *ménes- in av. maz-da- vergleiche, ist regelrecht, da PK im
Idg. zur Zeit der Schwundstufenbildung auf dem KE betont waren. Der Stamm *meges-
kann nun auf zweierlei Weise interpretiert werden: einmal als neutrales Abstraktum ‘Größe’,
belegt in ved. máhas-?? und av. mazah-; zum zweiten als ein daraus durch interne Deriva-
tion abgeleitetes Possessivadjektiv *megés- ‘Größe habend = groß’, vielleicht belegt in ved.
mahah, das an einigen Stellen am einfachsten als neutrales Adjektiv aufgefaßt wird (z.B.
RV 1.3.12 mahö arnah ‘große Flut’). Possessivadjektiva beliebiger Bildungsweise konnten
aber, wie ich an anderer Stelle zu zeigen versuchen werde, im Idg. formal nicht als KA fun-
gieren"! Diese Regel und der ohnehin zweifelhafte Status von *meges- machen es wahr-
scheinlich, daß *mgs- zum Abstraktum *meges- ‘Größe’ gehört. Ein idg. PK der Form
*mgs-X ist als ‘ein X habend, das mit Größe verschen (= groß) ist’ aufzulösen. Der Kompo-
sitionstyp ist der von ved. isu-hasta- ‘Pfeile in der Hand habend’, ursprünglich ‘eine Hand
habend, die mit Pfeilen versehen ist’. Hier könnte auch idg. *dus- ‘schlecht’ einzuordnen

346

sein, dem vielleicht ein Substantiv *déues- ‘Mange! zugrunde liegt’: *dus-X ‘ein X ha-
bend, das mit Mangel versehen (= mangelhaft) ist”.

Im gav. Adverb maš < *még-s möchte ich aus formalen Gründen am ehesten eine archaische
Nom.-Akk.-Form von *méges- ‘Größe’ sehen (vgl. *men-s in av. maz-da-), doch bin ich mir
über die syntaktische Berechtigung einer solchen Interpretation nicht im klaren (vgl. Brug-
mann, Grundriß 2.2.688). Als Adverb weiter verbreitet ist idg. *mégh, in ved. mahi, gr.
uéya und wenigstens indirekt aisl. miok.

Anmerkungen

a o B o t ne

12
13

Abzulehnen Darmesteter, MSL 2 (1875) 307 und Collitz, JAOS 51 (1931) 160ff., die aX.- mit gr.
é£ usw. gleichsetzen.

Humbach, Gathas 2.45f. stellt maY vermutungsweise zu jav. mo3u *alsbald', lat. mox.

Wenn im folgenden von mazänt- gesprochen wird, sind maz- und mas- stets miteinbegriffen.
aX.dànungmca, kasu.dänunamca ohne Var.

Zur Syntax s. Reichelt, Elementarbuch 258.

So nach Bartholomae u.a.; Lommel übersetzt, sachlich unangemessen, ‘zu den Getreidefeldern, die
reichlich Körner bringen und zu den Weiden, die wenig Körner bringen’.

Herzustellen wohl auch H. 2.9 statt des überlieferten amaiid huraoéaiid, doch lag amaiid schon dem
Pahlavi-Übersetzer vor, der das Wort mit ’m’wnd wiedergibt.

aXahunaram F1 Pt1 L18;aXahe.narom P13.

Der Passus ist verwandt mit Yt. 5.91 ha6a.hunaró tanu.mqóroó und Y. 57.33 taxmahe tanu.mq órahe
/. . .] bázuXaojaghó ra 9aextd.

aX.managha ohne Var. Es empfiehlt sich nicht, die überlieferte Form zu halten und als thematisiert
aufzufassen, da Thematisierungen von s-stämmigen PK praktisch nicht vorkommen. Deskriptiv am
besten bezeugt, wenngleich nur an minderwertigen Stellen, ist pouru.x’arananha- ‘viel Ruhmesglanz
habend': pouru.xYaranagho aYauua zara 9utró V. 19.3, pouru.xVaranagha ahura mazda (V ok.) Vyt.
24, yazata pouru.xVaronagha (Pl.) Ny. 3.11 u. 5.6 (auch Perseveration móglich), xl"aronó . . . pouru.
xVaronagham (Akk.) Yt. 18.1 (einfach die m. Form, wie das danebenstehende huX.hgm.baratam
statt ‘barat nahelegt). — Vgl. weiters V. 6.16, 18, 20, wo der ASg. astam ‘Knochen’ der Reihe nach
mit arazu.stauuayhom, parasu.masayham, * bi.orozu.stauuaghom, biporosu.masaghom, bàzu.stauuag-
ham, sraoni.masayham verbunden wird, bei denen es sich wohl gleichfalls lediglich um die m. For-
men handelt. -- Schließlich steht die Form stahrpaesayham nach Bartholomae s.v. an drei Stellen
bei Neutra. Zwei davon entfallen: Y. 57.21 ist nach Ausweis des Metrums mit Geldner nmanam ...
stohrpaësam zu lesen; Yt. 13.3 (asmanam) ... yim mazdäd vaste vayhanam /stahrpaesagham mainiiu.
taStam heißt nicht, wie man zu übersetzen pflegt, ‘(den Himmel,) ... den sich Mazda als sternenge-
schmücktes, geistgeschaffenes Kleid anlegt’, sondern vielmehr ‘den sich Mazdä als Kleid anlegt, den
sternengeschmückten, geistgeschaffenen'. Lediglich Y. 9.26 ist die Phrase stohrrpaesaghoam mainiiu.
täftam mechanisch auf das n. aifiidghanom übertragen. — Falsch beurteilt wurde von Bartholomae
Yt. 13.23 u. Par. frauuasaiiö aYoiX baeaza hacimnd zam.fradanha dànu.dràjagha huuara.barozanha,
wo natürlich, wie auch Kellens, Le verbe avestique (1984) 53 gesehen hat, von hacimnä abhängige
Instrumentale vorliegen.

Man beachte, daß stuui.kaofa-, das einzige andere Kompositum mit kaofa-, gleichfalls neben einem
aX.-Kompositum steht (af. bäzäuf, s. 2.2.).

Vel. II 487f.... radpor .. . [ aldwva peyadupov.

ax.varacó ohne Var.

Farhang-i öim. Edition und Kommentar (Diss. Erlangen 1968).

Ein altpers.-med. PN *A3-farnah- ist in der elam. Form AXparna bezeugt (dazu das Hypokoristikon
ASparnuka), s. Mayrhofer, Onomastica Persepolitana (1973) 130 mit Lit.

Vgl. RV 1.133.1 drüho dahámi sam mahir ‘Ich verbrenne zur Gänze die großen Lügengeister' (nach
Geldner).

Zum Nebeneinander von aX.- und stuui.- (so die ursprüngliche Schreibung) vgl. oben 2.2. aX.bázaàuY
stuui.kaofo.

BBS

347

Der Form nach beruht x$uuaößaiiat. wohl auf einem Transitivum *xSuuaéBaiia-, das Kompositum
wäre demnach (mit Bartholomae s.v. und Duchesne-Guillemin 201) synthet. NAg. ‘die Peitschen
schwingend’. Für die Figur ‘Reiche, die die Peitschen schwingen’ = ‘wo man die Peitschen schwingt’
sind mir allerdings keine Parallelen zur Hand. Andererseits wäre der Ansatz eines intransitiven
*xSuuaéfaiia- (so Lommel), der auf die gut passende Bedeutung ‘wo die Peitschen schwirren' führte,
morphologisch nicht ganz leicht zu rechtfertigen.

x34- wird bekanntlich sonst mit Gen. konstruiert. Hat der Kompilator unseres Stücks irgendwo
xSa dra xSaiiente, eigentlich ‘sie gebieten mit Macht’, vorgefunden und mißverstanden?

Kellens’ Konjekturen *a¥.uyra-, ta$.darazra- und implizit +a¥.viiéxanaiia-, taš.vərəðræyniia-, taš.
baèSaziia- (Yt. 14.46; ebenso schon Geldner, Drei Yasht aus dem Zendavesta, 1884, 83) bzw.
*aX.varaórauuan-, taX.amauuan- (Yt. 14.59), die er selber ‘sous toute réserve” vorschlägt (Stlr 4,
1975, 15421), dürfen auf sich beruhen. Statt af ist ds ‘war’ überliefert, und in der Form, in der uns die
beiden Stellen vorliegen (vgl. a.O. 149), ist jedenfalls die Kopula auch gemeint. Abzulehnen ist
Geldners (3Yt. 116) Erklárung des PN aYta.auruuant- als *aX.tauruuant-, s. Mayrhofer, Die avesti-
schen Namen (1977) 25f. mit Lit.

Vgl. Mayrhofer, Die altpersischen Namen (1979) 20f. mit Lit.

Das gleiche gilt für den elam -altpers. PN Pid(dJabarma, falls die geistreiche Deutung von Gershevitch,
Studia Pagliaro Il (1969) 223f. als *bida-barva- ‘seed-chewer’ (zu ved. bija- ‘Getreidekorn’) zutrifft.
Vgl. sügabhasti- RV 9.72.2 und jüngere, AiGr. 2.1.294 angeführte Beispiele. — Ein synthet. NAg.
dürfte durch den Akzent ausgeschlossen sein.

Falsch Bartholomae s.v. (‘der reichlich frühere’ oder ‘voranstehende’).

Unwahrscheinlich andere Versuche, mit aX.paouruua- zurechtzukommen, bei Geldner, 3Yt. 97, 103
(‘mit starken Fugen’, zu ved. párvan-) und Bailey, Zoroastrian Problems (1943) 6! (PK mit iran.
*parua- 'abundance', vgl. gr. rAobroc).

Yt. 19.9 aÉvandaram F1 usw., aYa.vandarom J10 D; Yt. 19.14 a&vandrom F1 usw., aXvandarom H3;
Yt. 19.45 aX vandrom F1 usw.

Für aX. vandra- z.B. auch Geldner, 3Yt. 11, Duchesne-Guillemin 131.

Kaum in geschichtlichem Zusammenhang mit dem Unàdis. 2.13 gelehrten ai. vandra-, Bed. nach
Ujjvaladatta ‘pùjaka-’.

Vgl. noch Klingenschmitt apud Mayrhofer, KEWA 3.568 Fn., der das KA von jav. xVandrakara-, etwa
‘schmeichelnd’ oder ‘gefällig’, als *hu-yandra- *wohllóblich' interpretiert. An den Belegstellen V.
13.46, 48 wird der Hund als xVandrakara- wie der Knecht bzw. die Hure bezeichnet. Die Auflösung
des Gesamtkompositums ist mir unklar (‘wohlgelobt machend”?).

Kellens a.O. 153 hält die Form seltsamerweise für einen Komparativ ‘beaucoup plus puissant”, ob-
wohl er 20jd und nicht das als v.l. (lectio facilior) gut bezeugte aojiid in den Text setzt. Der Kompa-
rativ von aX.aojah- hütte im Gav. gewif wie im Jav. as.aojastara- gelautet, s. 5.1.

aojah- als Adjektiv ist noch Yt. 10.140 belegt.

as J2.3.6.7.1;a$ nur im Indischen Videvdäd Säda (L1.2.3 S2 Bb1).

Hss. aojiftam, tanci$tam; vgl. zur Verderbnis Yt. 15.16 yat bauudni x’aranay’hastamam (F1 J10)
zätanam, wo das unabhängige K40 noch ‘tam hat.

as alle, nur J3 P11 ds; af Li; af L3 P1.

Yt. 10.98 as ohne Var. 10.135 as F1 Pt1 L18 P13 K15;2$ H4 M12; deest H3.

Geldner aXxrdxVanutamq, s. die Varianten.

afboratò Mf3 K13.38 HS; afa.baratò F1 usw. aSa.baratò J10.

zaoüa- ist im Jav. (außer im Wortsalat V. 1.14) nur Beiwort der FrauuaSis, vgl. noch Yt. 13.27 u.
148.

Die Bedeutung des KA von vazarat- ist unbekannt.

Synthet. NAg., die auf Verbalphrasen mit Pridikatsnomina im Nominativ beruhen, sind grundsätz-
lich bildbar, aber naturgemäß nicht háufig nachzuweisen. Vgl. unten 7 zu xYuuifii.vàza-.

Die Alternative, aX.barat- als ‘Großes bringend’ aufzufassen, entsprechend einem freien Syntagma
*mazät bar- o.ä., dürfte nicht zum Ziel führen. Es läßt sich nämlich nicht wahrscheinlich machen,
daß of. auch ein (ohnehin unbelegtes) Substantiv ‘das Große’ mitvertreten konnte: hu- und duš-

' haben jedenfalls nie die Funktion der neutralen Substantiva vohu- bzw. aka-. ‘Gutes wirkend’ heißt

42
43

vielmehr vohu-uaraz-, ‘das Schlechte schaffend' ako. dà-.
Herzfeld möchte *aspa.rt- ‘fighting on horseback’ oder *asa. brt- ‘riding on horses’ herstellen.
Daneben tuvisvani- wie av. stuui.- neben aS.-, s. Anm. 17.

348

44 Als Kompositionsformen von Adjektiven kònnen Adjektiv- und in bestimmten Fällen (einiges bei

50
51
52

Vf., FS Risch, 1986, 393ff.) Substantivstimme dienen; zu letzteren rechne ich mit Schulze apud
Fraenkel, KZ 42 (1909) 124? auch die KA mit Caland-i, worüber anderswo.

Vgl. AiGr. 2.1.66f., Duchesne-Guillemin 21f.

Entsprechend scheinen auch, worauf ich hier nicht eingehen kann, synthet. NAg. mit hu- nicht frei
bildbar gewesen zu sein.

Es sei hier angemerkt, daß sich Adjektivadverbia auf *-s nicht als eigenständige idg. Kategorie nach-
weisen lassen. Die beigebrachten Beispiele sind wohl in allen Fällen adverbialisierte Kasusformen,
s. Brugmann, Grundriß 2.2.677ff. (zu gav. ma$ s.u.).

Vgl. z.B. die &vdeos-Komposita.

Ein gutes Beispiel ist das von Bartholomae, IF 9, 283 seiner Bildung nach mit aX.- und gav. may ver-
glichene gav. Adverb ara¥ ‘recht, richtig’ (ursprünglich wohl NSgm. */3rg-s), das als KA in den syn-
thet. Komposita araZaji- ‘richtig lebend’, ərəžuxsa- “richtiges Sprechen’, jav. ar$.däta- ‘recht geschaf-
fen’ erscheint. Anstelle von *aro$.vac- ‘richtig sprechend’ finden wir jedoch ara, vacah-, zu dessen jav.
Entspechung ar¥.vacah- die parallelen ar¥.manah- ‘richtig denkend’ und or? Zoo äng. ‘richtig han-
delnd' hinzugebildet wurden. — M. Mayrhofer weist mich auf den elam.-altpers. PN Ma$-ka-ma
(Onom. Persep. 194 mit Lit.) hin, der wohl als *Maÿ-käma- ‘sehr verlangend” zu interpretieren ist.
Bildungen mit -Kkdma- konnten also nicht nur als PK, sondern auch als synthet. NAg. fungieren (KN).
Das -h- von mdhas- beruht auf Analogie nach dem Adj. mah- < *meg-h2-.

Zu denen mit *-went- und zum Typ ved. /suhasta- s. FS Risch 395ff.

Vgl. Eichner, GS Kronasser (1982) 2790,

The multiple origin of the Indo-European nominative case!

William R. Schmalstieg (The Pennsylvania State University)

From my early years as a graduate student at the University of Pennsylvania I have admired
Henry Hoenigswald, always one of my favorite teachers. I hope that it is not out of place
for me to praise not only his scholarship and keen understanding of historical linguistics,
but also his qualities of modesty and compassion as a human being. I feel deeply honored
to have the opportunity to contribute a small piece to this volume.

1. I have argued elsewhere that Indo-European was originally a language with split ergativity
(Schmalstieg 1980: 184—185). In general the forms of the middle aorist and perfect tense
were correlated with an ergative syntax and morphology, whereas the forms of the present
tense came to be correlated with an accusative syntax. The morphological correlation is
expressed by the relationship between the 3rd sg. middle aorist ending *-to, (cf. Gk. édo-
to, Old Indic ddi-ta ‘gave’) with the participle in *-tos, (cf. Gk. do-tós, Lat. da-tus, Lith.
duo-tas ‘given’, see Hirt 1928: 102 and Schmalstieg 1980: 93). Originally the participle in
*-to was completely intransitive as in Old Indic ga-td- ‘gone’. The oldest sentence type
would have been represented by (1) *reg gWm-to ‘the king has (is) gone’, (2) *reg ghWn-to
‘the king was (is) struck’. The latter sentence must be translated as a passive in English and
all of the Western European languages with which I am familiar. Nevertheless at this early
stage I interpret it as being without diathesis. The original phrase consisted of only two
elements, a subject and predicate which, if verbal, was intransitive. A third element, namely
an agent, could be added to such a sentence and one could have (3) “reg ghWn-t6 gWena-s
which we would be tempted to translate as ‘the king was struck by the woman’.”

This construction, however, is the original Indo-European ergative construction and the
passive interpretation only became possible with the appearance of the active voice as
outlined in paragraphs 2 and 3 below.®

2. At a later period of Indo-European there came to exist sentences of the following type:

(4) *patér ghWent vlkWom
father strikes at the wolf
(subject in (intransitive (indirect object in
absolutive case) verb) dative-locative case)

‘the father strikes at the wolf’. The subject *parer is in the absolutive case without any
marker, the verb *ghWent is intransitive and atelic and cannot take a direct object, but
rather takes the indirect object *v/kWom in the dative-locative case. The structure is a

350

typical antipassive construction known in various ergative languages, with a subject in the
absolutive case, an intransitive verb and an (indirect) object in an oblique case. (One notes
that in the BeZtinskij language of Daghestan the verb denoting ‘to strike’ is construed with
a dative object [Kibrik 1981: 3].)

3. Within Indo-European in a word-final sequence of short vowel plus sonant the sonant
was lost and the preceding vowel lengthened if the following word began with a consonant.
If the following word began with a vowel the etymological sequence was retained. Thus the
sandhi variants *pater, *vlkWom (prevocalic forms) and *paté, *vIk"© (preconsonantal
forms) came into existence. These original automatic variants came to contrast and the
meaning of absolutive case was ascribed to *paté whereas the meaning of the vocative case
was ascribed to *pdter (with the forward movement of the stress characteristic of this case
attested in Gk. pdter and Old Indic pitar). The form *paté is represented in the nominatives
singular: Old Indic pitd (with the root form *pi- [rather than *pa-] typical of Old Indic)
and Gk. patêr (with the addition of -r from other forms of the paradigm). The new form
*vyik"O retained the old dative-locative function (with the addition of -y [a]) and the old
form *v/kWom became the accusative case, viz., that case which could express a greater
degree of verbal transitivity than the dative. Thus sentence (4) is transformed into the
following sentence:

(S) *paté ghWent vlkWom
father strikes wolf
(subject in (transitive (direct object in
nominative case) verb) accusative case)

The creation of the new accusative case (by splitting from the old dative-locative) was oc-
casioned by an increase in transitivity. The reason for the increase in transitivity is as opa-
que as the reason for any syntactic change in any language.

Hopper and Thompson (1980: 252) have given a list of items which contribute to the
notion of transitivity. Thus a telic verb is high in transitivity; if the object is totally affected
it is higher in transitivity than if it is not affected. Timberlake (1977: 141—177) has given
an interesting account of an increase in transitivity in Russian in which the accusative case
has replaced the old (partitive) genitive as the direct object of negated verbs under certain
circumstances. Other examples of an increase in transitivity include the introduction of the
particle z- to denote a direct object in Armenian and the introduction of the ‘personal a’ in
Spanish.

According to Klimov (1973: 193) the origin of the accusative case is to be sought in the in-
direct case in many Kartvelian languages. In the language Udi, for example, the accusative
in -{a }x and -(e)x shows its origin in the dative case, but only in a reliquary fashion.

In Indo-European, the increase in transitivity led to the interpretation of the antipassive
(sentence 4) as an active (sentence 5). When sentence (5) became a true active then sen-
tence (3) could only be interpreted as passive.

4. Already Brugmann (1911: 122—123) had divided up the Indo-European nominatives
singular into (1) endingless forms, i.e., the -2, -(iJiià-: -I-, -n, r, -s stem forms and (2) forms
with the ending -s, i.e., the -o, -i, -u, -i, -4, dental, guttural (velar in our terms) stems and
root nouns.

351

As in my (1980: 183—185) work I still see the endingless forms as reflecting the old abso-
lutive case without any marking at all. Typically the subject case of an intransitive verb in
an ergative language is unmarked. The passage of the antipassive unmarked subject in the
absolutive case (typical of an ergative language) to the unmarked subject in the nominative
case (typical of an accusative language) is quite comprehensible.

The problem is to explain the -s of the nominative case. Earlier I had attempted to explain
this as the relic of a verbal-adjective formation derived from a stage prior to the separation
of adjective in (nom. sg.) *-os from the verb in *-es (31d sg. active < antipassive). Thus (6)
*patér bhor-os ‘father (is) a carrier, porter’ vs. (7) *patér bher-es ‘father (is) carrying’. The
*e/os denotes the animate pronoun whereas *-e/ot denotes the inanimate. In what came to
be the verb usually *-e? was generalized in the third singular, while *-es was generalized in
the second singular. Relics with *-s in the third singular are, however, encountered, e.g.,
Tokharian A 3rd pres. pälkäs ‘lightens’, Old Persian aiš ‘went’, Phrygian edaes ‘put’ (Schmal-
stieg 1980: 103).

The formal identity of the Hittite nom. and gen. sg. an-tu-uh-Sa-aS ‘man’ has led many to
identify the *o-stem genitive singular ending with an etymological ergative case. One
analyzes then Old Indic gen. sg. vir-as-ya as deriving from *vir-as plus a particle *io (Ivanov
1959: 28; MaZiulis 1970: 91) or a relative *vo (Lehmann 1981: 179-188). One could
suppose that a sentence such as (3) at the earliest period, may have had the agent *viros
‘by the man’. Thus:

(8) *reg ghWn-tö vir-0-s
king was struck by the man
(subject in absolutive- (intransitive (agent in ergative-genitive
nominative case) verb) case)

‘The king was struck by the man’.

5. Briefly digressing, note now the following old ergative sentence from Lithuanian (Paulaus-
kiené 1979: 105):

(9) Afit Sakéliu ant Zaliùju vainikai kabinta
On branches on green wreathes (are) hung
(nom. pl. (neut. sg. past
subject) passive participle)

*Wreathes are hung on the green branches’.

In modern Lithuanian the nom. pl. vainikai can be replaced by the acc. pl. vainiküs thus
changing the syntactic relationships and making vainikus the direct object of the verb.

The Lithuanian example furnishes a parallel with the Indo-European situation in which
sentences of type (8) with a nominative subject are replaced by sentences with an accusa-
tive object.* The passive interpretation of the verb was then replaced by an active inter-
pretation.?

(10) *res-m ghWn-tö Vir-o-s
king struck man
(direct object in (transitive (agent in genitive-ergative >
accusative case) verb) nominative case)

The man struck the king’.

352

Thus sentence (10) becomes parallel to sentence (5) and the agent in *-s becomes parallel
to the absolutive case without any *-s. The *-s spread in the nominative case to the extent
that it was felt necessary to mark the agentivity of the nominative case. Suffixes with high
agentivity such as *-a, *-ter, etc. did not require the *-s.

The penetration of the ergative into the absolutive case function is known, e.g., in western
Georgian dialects (Boeder 1979: 443). Here one encounters k ac-ma movida ‘the man
came’ and k ac-ma mok vda ‘the man died’ where the ergative ending -ma has replaced the
nominative -i which standard Georgian uses for the subject of such intransitive verbs.

I suggest furthermore that the introduction of the *-s into the nominative case was an
ongoing process which was never, indeed, completed. One notes that in Hittite the personal
names are sometimes supplied with a final -š and sometimes not. Thus one encounters the
nom. sg. la-ba-ar-na-aš ‘king’ (deriving from an old proper name) as well as /a-ba-ar-na (Gii-
terbock and Hoffner 1980: 41) without +, similarly Mur-Si-li and Mur-Xi-li-i8, etc. As Čiko-
bava (1948: 223) has pointed out, in Old Georgian the indefinite case was originally used
for agent with proper names and one encounters Abraam $va Jsaak ‘Abraham bore Isaac’
with Abraam in the indefinite case rather than the expected ergative Abraam-man. On
Silverstein's agentivity scale the proper nouns are more likely to function as a transitive
agent than common nouns, so the marking with the *-s was later in the proper names
(Dixon 1979: 85).

Schwyzer (1966: 64) writes that in Greek the inanimate masculines, feminines and neuters
were limited to the function of subject and predicate in nominal sentences and to the func-
tion of subject of stative-fientive intransitives and passives, whereas animate masculines,
feminines and neuters (thus zeknon ‘child’ or collectives like stráteuma “army, host’) could
function as the subject of verbs of action (the so-called energetica), which were mostly
transitive. Tchekhoff (1978) quoting Laroche (1962) and Benveniste (1962) has noted that
in Hittite an inanimate noun can be the subject of an intransitive verb or the object of a
transitive verb, but not the subject of a transitive verb. Thus, for example, the neuter word
pahhur ‘fire’ can be the subject of the intransitive verb warani ‘burns’. On the other hand,
in order for the noun ‘fire’ to become the subject of a transitive verb it must take the
ending -5, e.g.:

[man] antuhffan] huwahhlurtin] plah]huenaÿ epzi
when man throat fire seizes
(acc. sg.) (acc. sg.) (nom. sg.)

‘When fire seizes a man by the throat' (Tchekhoff 1978: 231).

There is some doubt as to whether this is a nominative or a genitive form, but even if it is a
genitive form, it would illustrate the principle that the more agentive nouns rendered the
active voice more transitive.

Thus the Indo-European endingless nominative cases derive from the old absolutive case
and the sigmatic nominative derives from the old ergative (or perhaps, if my earlier notion
is correct) from the old intransitive adjective-verb. The difference in the syntactic relation-
ships represented in the active voice (< antipassive) penetrated into the middle and perfect
tenses and Indo-European shifted from a language with split ergativity to an almost com-
pletely accusative language (remnants of the old ergativity only being retained in partici-
pial constructions in *-/o in Baltic, Slavic, Old Indic, Greek, etc.). The endings of the

353

middle voice and perfect tense are much more difficult to reconstruct than those of the
active voice. The reason is clear. Although the middle voice and perfect tense are historical-
ly older than the active voice, their syntax and morphology have been completely reshaped
by the latter.

Notes

I would like to thank Philip Baldi for suggestions and references, particularly in regard to the classical
languages. Naturally I assume full responsibility for the contents of the paper.

In Lithuanian passive constructions the genitive case denotes agent as opposed to the instrumental
which denotes instrument, so that an expression such as parak~ta mano ‘written by me (gen.)’
contrasts with paraxyta pieXtukü *written with a pencil (inst.)’. One can assume an original ergative-
genitive agent in Indo-European also, cf., e.g, Gk. Dids-dotos ‘given by Zeus’ and for Old Indic see
footnote 6 below. Further evidence is given in Schmalstieg 1980: 176—178.

Indeed sometimes it is difficult to make out the interpretation in Greek. Thus Schwyzer (1966:
237) writes that the phrase (N 597) to d’ephelketo meilinon egkhos could be interpreted either as
‘the ashen spear dragged after’ (der eschene Speer schleifte nach) or ‘was dragged after’ (wurde nach-
geschleift), or even ‘he dragged the ashen spear after himself’ (er schleifte den e. Sp. hinter sich
nach). Likewise (A 5) Diós d' eteleieto boulé could mean ‘the will of Zeus was going towards fulfil-
ment’ (ging der Vollendung entgegen) or ‘was being fulfilled’ (vollendete sich or wurde vollendet).
Such examples as the following are usually explained as being accusative of relationship, but I pro-
pose that they could be explained as reflecting the activization of a passive construction, i.e., the
substitution of an accusative for a nominative. From the Iliad: (PS 777) en d'ónthou boéou pléto
stoma te hrinas ‘and with the filth of bulls were his mouth and nostrils filled’. (E 354) melaineto de
khröa kalon ‘and (her) fair flesh was darkened’. The Greek examples fit exactly with the proposed
theory and one can compare the relationships in (PS 777) with those of Old Armenian as explained
below in footnote (5) by Benveniste. Note even the genitive case of ónthou boéou which originally
functioned as ergative agent. In Latin such examples include particularly verbs denoting ‘putting on
or taking off clothing, adorning, crowning, covering, etc.’ (Kiihner-Stegmann 1962: 288-289). Thus
(Ov. M. 1, 270) varios induta colores ‘clad in robes of many hues’, (Verg. E. 6, 68) floribus atque
apio crines ornatus '(his) hair adorned with flowers and with bitter parsley', (Verg. A. 12, 416)
faciem circumdata nimbo *(her) face veiled in dim mist’, (Verg. A. 8, 457) tunicáque inducitur artus
‘and (he) clothes his limbs with a tunic’. Kühner-Stegmann (1962: 285) write that the Latin accusa-
tive of relationship must be under Greek influence since it only appears in Augustan time in Latin,
originally in poetic language, particularly in Virgil and Ovid and in prose only in Tacitus. I would not
care to give an opinion on all the examples, but I see no reason why the same syntactic forces could
not have been at work in Latin as in Lithuanian (and apparently Greek), namely the substitution of
an accusative for an old nominative. See also Gotab (1975: 15).

A somewhat similar phenomenon is observed in Old Armenian where the accusative replaces the old
nominative. Benveniste (1971: 159) writes “In other words, z-gorc gorceal e nora ‘he has accom-
plished the work’ does not mean ‘eius facta est opera’ but ‘eius factum est operam.’ Since ‘eius
factum est’ is the equivalent of ‘habet factum’, it is not surprising that ‘eius factum est’ adopts the
transitive government of the earlier fecit, which it replaces in Armenian, and that it requires a
determined object.”

If, however, the *-to of the verb was used as a participle and there was verbal agreement between the
subject and the participle the agent remained in the genitive case and the syntactic change did not
take place, cf. Old Indic:

patyuh krita sati
by the spouse bought wife
(gen. sg. (fem. nom. sg. (nom. sg. fem.

agent) past passive participle) subject)

354
References

Benveniste, Emile, 1962. “Les substantifs en -ant- du hittite”, Bulletin de la Société de Linguistique de
Paris 57: 44—51.

Benveniste, Emile, 1971. “The passive construction of the transitive perfect", (Original in Bulletin de la
société de linguistique de Paris 48: 52-62) Problems in general linguistics (English translation by
Mary Elizabeth Meek) (Coral Gables: Univ. of Miami press), 154-161.

Boeder, W., 1979. “Ergative syntax and morphology in language change: The South Caucasian lan-
guages”, in: Ergativity, edited by Frans Plank (London, New York: Academic press), 435—480.

Brugmann, Karl, 1911. Grundriß der vergleichenden Grammatik der indogermanischen Sprachen V ol. 2,

„ Part 2 (Straßburg: Karl J. Trübner).

Cikobava, Arn., 1948. *"Istorifeskoe vzaimootnosenie nominativnoj i érgativnoj konstrukcii po dannym
drevne-gruzinskogo jazyka", in: Izvestija AN SSSR. Otdelenie literatury i jazyka. 8, No. 3, 221—234.

Dixon, R.M.W., 1979. “Ergativity””, Language 55: 59—138.

Gotab, Zbigniew, 1975. “Endocentricity and endocentrization of verbal predicates: illustrated with
Latin and Slavic material”, General Linguistics 15: 1—35.

Giiterbock, Hans G./Harry A. Hoffner, 1980. The Hittite Dictionary. Vol. 3, Fasc, 1 (Chicago: Oriental
Institute of the University of Chicago).

Hirt, Hermann, 1928. Indogermanische Grammatik Pt. 4. Doppelung Zusammensetzung Verbum (Hei-
delberg: Carl Winter).

Hopper, P.J./S.A. Thompson, 1980. ‘‘Transitivity in grammar and discourse”, Language 56: 251-299.

Ivanov, V.V., 1959. Toxarskie jazyki (Moscow: Izdatel’stvo inostrannoj literatury).

Kibrik, A.E., 1981. Materialy k tipologii érgativnosti: 13. Be%tinskij jazyk (Predvaritel’nye publikacii:
Fascicle 140) (Moscow: Institut russkogo jazyka AN SSSR).

Klimov, G., 1973. Oterk ob&ej teorii érgativnosti (Moscow: Nauka).

Kühner, Raphael/Carl Stegmann, 1962. Ausführliche Grammatik der lateinischen Sprache Part 1, 4th
ed. (Munich: Max Hueber Verlag).

Laroche, E., 1962. “Un ‘ergatif’ en indo-européen d’Asie Mineure”, Bulletin de la société de linguistique
de Paris 57: 23-43.

Lehmann, Winfred P., 1981. The genitive singular ending in -syo: How an Indo-Europeanist works”, in:
Bono Homini Donum: Essays in Historical Linguistics in Memory of J. Alexander Kerns, edited by
Arbeitmann, Yoél L. and Allan R. Bomhard (Amsterdam: John Benjamins B.V.), 179—188.

Ma£iulis, Vytautas, 1970. Baltu ir kitu indoeuropietiu kalbu santykiai: Deklinacija (Vilnius: Mintis).

Paulauskienė, A., 1979. Gramatinés lietuviu kalbos veiksmatodio kategorijos (Vilnius: Mokslas).

Schmalstieg, William R., 1980. Indo-European linguistics: A new synthesis (University Park and Lon-
don: The Pennsylvania State University Press).

Schwyzer, Eduard, 1966. Griechische Grammatik Vol. 2. (Munich: C.H. Beck).

Tchekhoff, Claude, 1978. "Le double cas-sujet des inanimés: un archaisme de la syntaxe hittite”,
Bulletin de la société de linguistique de Paris 713: 225—241.

Timberlake, Alan, 1977. "Reanalysis and actualization in syntactic change”, in: Mechanisms of syn-
tactic change, edited by Charles N. Li (Austin and London: University of Texas Press), 141—177.

An Indo-Iranian etymological kaleidoscope

Hanns-Peter Schmidt (University of California, Los Angeles)

1. Sanskrit darad ‘heart’

In his commentary on Unadisùtra I 129 Ujjvaladatta gives ‘heart’ as one of the meanings of
the word darad. Since there is no literary attestation, it is of course impossible to tell
whether the meaning is correctly reported or based on a misunderstanding of the source.
The word is listed as a derivative of the root dy ‘to split, tear asunder’. Semantically this
etymology is implausible. If the word is real at all, it is possibly of foreign origin. Thus the
Old Persian word for ‘heart’, *dorod, suggests itself. If this is correct, the borrowing oc-
curred at a time when the change to Middle Persian di/ had not yet taken place. The ulti-
mate literary source from which Ujjvaladatta's information stems may then be 1500 or
more years earlier than he himself (end of 13th century A.D.).

2. Sanskrit darada ‘red lead’

The shorter Petrograd dictionary quotes darada n. from the Bhävaprakäsa in the meaning
‘red lead’. A derivation from dr is unlikely. A colour term would be a more probable can-
didate. In genuine Old Persian we have daraniya ‘gold’ and would expect *darita as the
equivalent of Avestan zairita ‘yellow, yellowish red’. The colour range covered by deriva-
tions of the base zar extends from yellow over green to red (cf. Bailey 1974: 372). Firdosi
used ab i zard of blood (cf. Lentz 1926: 312). In Middle and New Persian we find only
initial z in the derivations of the base zar. The genuine Persian forms, *dar etc., were re-
placed by the common Iranian ones, possibly because of the numerous homophonic clashes.
The continuation of *darita, *dard, clashed with dard ‘pain’. It is however possible that
the old South Western form was partially retained, either dialectally or in a specialized
term.

3. Sanskrit kudaka ‘child’

In Sanskrit the word kudaka is so far attested only once in the Kashmiri tradition of the
Mahäbhärata 6.2.8. The correct meaning, ‘child’, was recognized by Burrow (1957: 6) and
Agrawala (1959: 30). Burrow based his interpretation on the North Western Prakrit kudaga,
fem. kudi, of the so-called Niya documents. At one time the word must have had wider
currency in Sanskrit since the Dhatupatha VI 89 lists a verb Kudáti in the meaning 5balye

356

‘acting like a child’. In his glossary (1937: 83) Burrow stated: “A connection with N. Pers.
kudak ‘child’ is not out of the question.” Later he preferred to derive the word from
Dravidian (1943: 135 n. 3). Tamil has kura ‘young’, Kanarese koda ‘youth’. In Munda lan-
guages homophonic words also occur, thus Santali kora ‘boy’, kuri ‘girl’. Burrow suggested
that the Munda words derive from Dravidian through Indo-Aryan mediation. Turner
(1966: no. 3245) leaves it open whether the ultimate origin is Munda or Dravidian. With
the sole exception of Bengali df-kurd ‘childless’ the possible Modern Indo-Aryan cognates
listed by Turner are restricted to North Western India, and this fact makes Dravidian or
Munda origin somewhat doubtful. The Iranian connection should therefore be reconsidered.

From Niya Prakrit dhane < Iran. *danaka ‘4th part of the drachma’, ajhate < *azataka
‘free’ and saste < *sastaka ‘day’ Burrow (1935: 783; cf. 1937: 99) concluded: “It appears
that in Iranian words original zka is represented by -e in this language." If this were a strict
rule, one would be tempted to infer from it that in the Middle Iranian source language final
g X k was already dropped. This would however be a fallacy since the development -aka >
-aga > -ae > -e is also otherwise attested in Niya Prakrit. Burrow (1937: $ 53) mentions the
last stage in the discussion of the declension, but omits it in the phonology ($ 8; 16).
Burrow’s ‘rule’ can consequently not be used as an argument against the possible Iranian
origin of kudaga.

The Iranian form will have been *kuóak or *kudag. The vocalism of the first syllable agrees
with Avestan kutaka ‘small’. Pahlavi kwtk, Manichaean Middle Persian gwdk and New
Persian kodak ‘young, boy’ show guna-forms. It cannot be determined whether the final
consonant was k or g since we do not know whether the source dialect had preserved the
tenuis of the diminutive suffix as New Persian did (cf. Horn 1898: 64).

The Iranian words derive, as Bailey (1954: 144f.) has shown, from an Indo-European root
*kau : ku ‘to be small’.

4. Rgvedic kumaradesna

Bailey (1954: 144f.) analyzes kumara ‘boy’ as ku-ma-ara from the root ku ‘to be small’,
and Burrow (1955: 187f.) accepts this explanation. It is much superior to Bóhtlingk's
etymology, ‘easily dying > perishable > tender’. Bòhtlingk found the only adjectival use of
kumard in the bahuvrihi kumarddesna (Rgveda 10.34.7), an epithet of the dice. Since the
gains one makes with dice do not perish but are rather lost again, the interpretation ‘whose
gifts are perishable’ is not very plausible, and Geldner proposed in his glossary the transla-
tion ‘whose gifts are those of children’, i.e. the gifts once given are taken away again. As
Oldenberg emphasized, the phrase following the word in the text strongly supports this
interpretation: jdyatah punarhdnah ‘beating again the winning one.’ Neisser (1930: 61f.)
argued that claiming gifts back is not a characteristic behaviour of children, and he defended
Böhtlingk’s translation. Burrow maintains that Geldner’s interpretation presupposes an
abnormal bahuvrihi, and he prefers ‘whose gifts are small’. There is however nothing wrong
with Geldner’s dissolution of the compound (cf. Wackernagel-Debrunner 2/1 § 101 for
parallels). Moreover, in the whole history of Sanskrit, Pali and Prakrit kumara is always a
substantive, never an adjective, which agrees with the use of the suffix -dra (Wackernagel-
Debrunner 2/2 8 174). As to Neisser's criticism, based on the opinion of a psychologist, it
may be said that even though taking gifts away again is not habitual with children it is of
frequent enough occurrence to make it appear to be a characteristic trait.

357
5. Rgvedic kavatnu, kavasakhd, kavari etc.

Hoffmann (1957/1976: 411f.) has convincingly argued that kavatnu presupposes a present
stem *kavati and that kavasakha and kavari are compounds of the type siksanara ‘giving
gifts to men’. kavdsakhá occurs as an antonym of maghdvan ‘generous, liberal’ in Rgveda
5.34.3, and the meaning generally assigned to it, ‘stingy towards the companion’, is plausi-
ble. The kavari in 10.107.3 are characterized by the words nahi te prnänti ‘for they do not
give gifts’, which clearly defines them as ‘stingy to the stranger’. This is acknowledged by
Hoffmann. It should be added that in 7.32 the pada 9d nd devasah kavatndve stands in
obvious opposition to pada 8d prndnn it prnaté mdyah ‘only the giving one is a joy to the
giving one’ which determines the meaning of 9d as ‘the gods (are) not for (= in favour of)
the stingy one’.

In his search for an Indo-European root which can be connected with the reconstructed
Vedic *kavati Hoffmann finds *kau ‘to humiliate; disgrace; shame’ most suitable. This root
has been posited on the basis of Germanic and Baltic material. Hoffmann wants to inter-
pret the Vedic words accordingly. Although it is conceivable that one humiliates somebody
by stinginess, it is not very plausible to assume that terms which literally express the humil-
iation acquire the sense of stinginess.

Hoffmann became acquainted with Bailey’s and Burrow’s reconstruction of a root *kau : ku
‘to be small’ only after he had formed his own hypothesis. He concedes that the Indo-
European *kau appears to have had the concrete meaning ‘to make small, inferior’, but by
restricting it to the factitive sense he demonstrates his bias. He is also at pains to stress that
Pahlavi kwtk ‘small’ has also the sense lowly, humble’ and thus means the same as Gothic
hauns ‘lowly’. Furthermore he recognizes with Herzfeld Old Persian skau@i ‘lowly, poor’
and Pahlavi %kwh ‘poor’ (vrddhi-formation of *sku-@a) as cognates of kwtk etc. In Pahlavi
kw'ifk) ‘young (of man or animal)’ he finds the substantive kava which is also present in
Vedic dkava. While this is formally unobjectionable, semantically the interpretation of kava
as ‘Erniedrigung, Beeinträchtigung, Minderung’ is unconvincing. Only in very special
contexts would one speak of the young as being characterized by ‘diminution’ rather than
simply by 'smallness'.

Deficient, lowly, humble, poor, miserable, despised' are universally attested secondary
connotations of words for ‘small’. ‘Disgrace, shame’ are also easily derivable from a basic
meaning ‘to be, become, feel small’. Only Gothic haunjan ‘to make lowly, humiliate,
degrade’ and Old High German honen ‘to scorn’ show factitive meaning which is however
best explained by the fact that these verbs are denominatives. That Germanic and possibly
also Greek (if kaüpoc ‘kakôc belongs here) have preserved only pejorative connotations, is
a special development. In Baltic, Lettish kauns ‘disgrace, shame’, Lithuanian kuvétis ‘to be
ashamed’ show the pejorative sense while Lithuanian kumelys ‘foal’ and kuméle ‘filly’
continue the concrete meaning ‘to be small’. In Indo-Iranian there is no trace of the facti-
tive sense ‘to humiliate’. The words kavatnu, kavasakhd and kavari express stinginess which
is frequently associated with smallness as demonstrable by English petty and German klein-
lich.

Since humiliation does not belong to the connotations of the root ku in Indo-Iranian, Hoff-
mann's suggestion (apud Mayrhofer 1953—1980: III 677) to explain the Old Persian name
Kuru as ‘humiliator of the enemy in the verbal contest" must also be rejected. The evidence

358

Werba (1980: 16ff.) adduces from Xenophon’s Cyropaedia in support of Hoffmann’s
hypothesis is fallacious: it would have probative force only if it showed that the name of
Cyrus was associated with bragging and singing the battle-song. Mayrhofer (1979: II 23)
is right in leaving the question of the etymology of Kuru open since even the nowadays
most widely advocated connection with the root ku ‘to be small, young’ (cf. esp. Sze-
merényi 1977: 13ff.) lacks the corroboration required. Incidentally it may be mentioned
that Wiist’s (1966: 77) attempt to show that ‘young dog’ is the original meaning of Kuru
meets with the same objection as Werba’s argument: our sources do not associate the name
of Cyrus with his having been nursed by a dog.

With Hoffmann Vedic dkava is best explained as a bahuvrihi. Its literal meaning is ‘without
smallness, deficiency’, and it is used as a litotes; Geldner’s translation ‘vollkommen’ = ‘per-
fect’ is quite adequate. Hoffmann also explains Kava in kdvatiryafic ‘directed a little across’
convincingly as an instrumental of the substantive kava.

Burrow (1955: 189ff.) also places here the common prefix ku which was generally regarded
as a form of the interrogative pronoun (Wackernagel-Debrunner 2/1: 82) because from the
later Vedic language onwards it alternates with other forms of the interrogative (ka, kad,
kim). Panini took the pejorative sense of ku as the basic one as can be inferred from his
sutras 6.3. 105ff. where he teaches the substitution of ku by other forms when the sense
isat ‘a little’ is intended. Historically the development was the other way round: ku ‘little’
also assumed a pejorative sense which became predominant though not exclusive. According
to Burrow the interpretation of ku as an interrogative is secondary and has triggered the
alternation with the other forms. Burrow states that outside Indo-Aryan such a use of the
forms of the interrogative pronoun is not attested. He does not discuss the parallels which
have actually been adduced.

Obviously he rejects, as Charpentier (1924: 14) had done before him, the parallels which
Bartholomae believed to have found in Avestan (cf. the complete list in Duchesne-Guille-
min 1936: $ 183). In Avestan ku.ndiri, generally translated by ‘harlot’, Burrow finds the
pejorative sense of ‘small’. This is plausible. It should however be added that the word is, in
its only attestation, used of a homosexual (Videvdàd 8.31—32) and probably had not the
sense of ‘harlot’, but that of the partner acting the female part in homosexual relationships,
and consequently the word is best understood as a ‘deficient woman” because the homo-
sexual is not a real woman.

There is another word which may contain ku: kuruya. If this is the name of a disease,
which is by no means certain, it could be connected with Sanskrit röga ‘disease’, as Schwy-
zer (1929: 113) suggested. Bad disease’ would be suitable, but it is uncertain whether the
existence of the root rug/ruj can be presupposed in Iranian (cf. Mayrhofer 1953—1980:
IIl 64 s.v. rujáti). On the Avestan words with initial ka see the following note.

The most striking and persistently quoted parallel from outside Indo-Iranian is the Boeotian
novAuos (Old Boeotian ruAuoc) ‘awful hunger’ which Schulze (1895/1933: 399f.) sepa-
rated from Greek BovAywos for phonetic reasons. For ßouAtuos and other compounds with
Bov as intensifier he accepted the connection with Bovs because the Greeks themselves
recognized it. He interpreted the Boeotian prefix mu as the equivalent of Sanskrit ku fol-
lowing the interpretation of the Petrograd dictionary according to which it originally had
the function of emphasizing the extraordinary, unusual aspect of a phenomenon. He deemed
this isolated example as sufficient evidence to claim Indo-European age for this peculiar

339

type of compound. There have been few to take issue with this ingenuous but daring asser-
tion. Meillet (1916: 349) said that there is no reason to separate BovAyos from zo Auge
because in Slavic the alternation between ka and ga occurs in similar compounds (see also
the next note). This is a perverse argument since alternations in one language family cannot
prove anything for alternations in another family. From the context one must however
infer that Meillet approved of Schulze’s etymology and rather wanted to separate the
prefix fou from the name of the bovine. Bechtel (1921: 246f.) pointed out that one would
expect rather kv, kov in Boeotian, but accepted Schulze's interpretation in spite of this
serious obstacle. Charpentier (1924: 14 n.1) rejected it without giving any reason. Chan-
traine (1968ff.: 1 187) still quoted it with approval. In my opinion the phonetic difficulty
is serious enough to reject it. If the alternation between Greek Bou and Boeotian mov cannot
be justified by any analogies, another derivation of the Boeotian word may be considered:
the prefix mv could be connected with the adverb mv, attested in the superlative tUuuaros
‘extreme’. This suggestion is offered for what it is worth: the origin and exact meaning of
the adverb zv are after all not definitely established (cf. Frisk 1952—1970: II 624).

Other alleged parallels to the use of the interrogative in Sanskrit compounds are from
modern languages and can hardly be claimed as proof for the Indo-European age of the
compound type. Thus we find in Byelorussian ¢ymalyi ‘rather big, considerable’ “is that
supposed to be small?!” (Fraenkel 1923: 243). Such words probably belong to those
‘Augenblicksbildungen’ which happen to attain wider currency like the English whatnot.

6. Avestan kamaraéa and the problem of the prefix ka

Taking it for granted that the Avestan word for the head of a daeévic, demonic creature,
kamarada, is related to Sanskrit murdhdn ‘head’, Bartholomae (1904: 440) explained the
prefix ka as the interrogative in a pejorative sense: ‘what a head” = ‘a hideous head’. The
other examples which Bartholomae claims for this type are all words of negative sense:
kaxvarada ‘magician’, fem. kaxvaraidi (cf. the loanwords in Armenian kaxard ‘YOng,
vappuakeUc, Sanskrit kakhorda, kakkhorda, kharkhoda, kharkhota ‘a kind of magic’;
possibly also Niya Prakrit *khakhorda according to Burrow (1935: 781; 1937: 86));
kaxuti ‘an evil female’; Xapasti ‘a certain infectious disease’. Bartholomae’s etymologies of
these words are far from convincing. To connect kaxvarada with Gothic swarts ‘black’ is a
hazardous guess since relatives of the Gothic word are lacking in Iranian. The same objec-
tion applies to the collocation of kapasti with Latin pestis which is of unknown origin.
kaxuZi could be related to Sanskrit kuhaka ‘trickster, juggler’, but the latter word can also
be derived from kuha ‘where’ (Mayrhofer 1953—1980: I 249).

Schwyzer (1929: 111f.) added kayaôa to the list. He assumed a borrowing from Old Per-
sian which has yad ‘to worship’ against Avestan yaz. kayaa would then mean ‘how (= bad-
ly) worshipping’ and by the dialectal form also allude to the pronunciation of the people
performing the outlawed worship. In spite of Duchesne-Guillemin’s attempt (1936: $ 183)
to support this hypothesis, it has little to recommend itself. There is no evidence which
allows us to single the Persians out as arch-devil-worshippers. The relation between kayada
and kaiöya does not favour this etymology (cf. now also Dehghan 1982: 70 for the sug-
gestion to interpret kaióya as gen. sg. of the fem. Kkayeiói). An Old Persian word kayädä,
which is doubtlessly related, has become known in the meantime (Steve 1975: 10f.). It

360

occurs in juxtaposition with vatum (= Avestan yatu ‘sorcery’) and is probably a neuter
plural form (differently Mayrhofer 1978: 30f.). The Akkadian version has: ‘may (the gods)
neither damage nor destroy what I have built’. In the Old Persian version the verb is not
preserved, only the initial vi and the imperative ending -itu. Probably a verb meaning ‘to
meet with, fall into’ has to be restored: ‘let what I have built not meet with damaging
sorcery, not with destruction’. Although we cannot expect an exact translation in the
Akkadian version, it shows that the destructive force of yaru and kayada was felt as domi-
nant. This agrees with the Pahlavi translation of Avestan kayada : kastar ‘diminisher, de-
stroyer'. Also kaxvaraóa and kaxuzi were interpreted as kastar, but specifically of xvarrah
‘fortune’.

In all probability kayada has to be dropped from the list of words possibly containing the
prefix ka. If the other words already had a negative sense in their uncompounded forms,
the pejorative meaning of ka cannot be ascertained.

This must be emphasized also in view of the sweeping assertions of Meillet (1916: 348f.)
who accepted Bartholomae's theory without reservations. As to Kamarada he went even so
far as to maintain that this Avestan word with its short root vowel, contrasting with the
long vowel in Sanskrit mürdhan, suffices to show that the constitution of two different
vocabularies for good and evil beings is very ancient because the loss of the (laryngeal) a
(which caused the length in murdhän) in noninitial position goes back to high antiquity in
Iranian. Pahlavi kamal < *kamarda presupposes the existence of the uncompounded word
in Iranian. Furthermore Meillet adopts the theory of Miklosich that a Slavic word for the
raven, Serbian kdvran etc., contrasting with Old Slavic vraná, is a compound of the type
like Avestan kamarada. He does not spare a word about the question of whether Serbian
kävran is an evil raven and Old Slavic vranù a good raven. There is a distinction between
voron ‘raven’ and kdvoron ‘rook’ in Russian, but it is certainly not one of good and evil.
Even if the purely linguistic part of Meillet's argument is accepted, the conclusions as to
the differentiation of the vocabulary are more than hazardous.

Charpentier (1924: 12ff.) rejected Bartholomae's theory and proposed kamara$a with
Sanskrit kamatha ‘tortoise; begging-bowl of an ascetic'. Although semantically there is no
difficulty in associating the shell of the tortoise with the skull — as Latin testa > French
téte and Sanskrit kepala show —, there remain phonetic obstacles which can be overcome
only by complicated hypotheses.

A weighty argument against Bartholomae’s theory can be derived from Khotanese kamala
(< *kamrda) ‘head’ in a non-pejorative sense. From the Pamir dialects Bailey (1979: 52)
quotes Yidya kyemadyo ‘skull’. The differentiation of daévic and ahuric vocabularies in
the Avesta is not a homogeneous system. In the case of kamoroóa it is not possible to tell
whether it already had an exclusively pejorative sense when the religious differentiation
was introduced. Its ahuric counterpart, vay5ana, is of unknown etymology and has no
certain cognates in other languages. It is hardly a learned word as Bartholomae thought
since then one would expect rather a term of transparent etymology as we have it in
paitixtana ‘leg’. I would prefer to see in it an archaic word like staman ‘mouth’. kamarabda
may have had the restricted sense of ‘skull’ and thus offered itself readily to demonization.
There is another word for ‘head’ in Avestan, sarah; it was probably the vox media because
it continued in this use in several Iranian dialects, among them Persian.

361

I am quite aware of the fact that negative terms can undergo an amelioration. That this was
the case with Khotanese kamala, is however a rather remote possibility.

Pisani (1954: 125f.) draws attention to Bartholomae’s earlier proposal (1895: 224ff. cf.
also Bloomfield 1895: 423) to explain kamarada as a “proportional formation" to the
Iranian equivalent of Sanskrit kakúbh ‘peak’ (Latin cacumen) if this can be connected with
Cretan kupn ‘head’. Pisani picks up Fick's collocation of murdhän with Greek uéXadpov
‘ceiling, roof, house’ and sees in the variant kué\edpov the etymological equivalent of
kamarada whose a he explains as a mere anaptyctic vowel. k is for him a velar ‘praeformans’
also found in Slavic kosti ‘bone’, Latin costa ‘rib’ besides os ‘bone’ and other words. While
in some cases the praeformans is accompanied by a change of meaning, in others it is not.
It remains an enigma.

As befits the Ahrimanian pate, kamoraóa will continue to confuse scholars.

References

Agrawala, V.S., 1959. “Notes on Sanskrit Words”, Journal of the American Oriental Society 79: 30.

Aufrecht, Th. (ed.), 1859. Ujjvaladatta s Commentary on the Unädisütras (Bonn: Adolph Marcus).

Bailey, Harold Walter, 1954. “Indo-Iranian Studies’, Transactions of the Philological Society 1954:
129—156.

Bailey, Harold Walter, 1974. “The Range of the Colour zar- in Khotan Saka Texts", Mémorial Jean de
Menasce (Louvain: Imprimerie Orientaliste), 369-374.

Bailey, Harold Walter, 1979. Dictionary of Khotan Saka (Cambridge: University Press).

Bartholomae, Christian, 1895. “Arica VI”, Indogermanische Forschungen 5: 215—227.

Bartholomae, Christian, 1904. Altiranisches Wörterbuch (Straßburg: Trübner).

Bechtel, Friedrich, 1921. Die Griechischen Dialekte 1 (Berlin: Weidmann).

Bloomfield, Maurice, 1895. “Assimilation und Adaptation in Congeneric Classes of Words”, American
Journal of Philology 16: 409—434.

Burrow, Thomas, 1935. “Iranian Words in the Kharosthi Documents from Chinese Turkestan IP”,
Bulletin of the School of Oriental Studies 7: 779—790.

Burrow, Thomas, 1937. The Language of the Kharosthi Documents from Chinese Turkestan (Cam-
bridge: University Press).

Burrow, Thomas, 1943. “Dravidian Studies HI”, Bulletin of the School of Oriental and African Studies
11: 122-139.

Burrow, Thomas, 1955. “Sanskrit kava- and Related Words”, Indian Linguistics 16: 186—193.

Burrow, Thomas, 1957. “Sanskrit Lexicographical Notes”, S.K. Belvalkar Felicitation Volume (Delhi:
Banarsidass), 3-11.

Chantraine, Pierre, 1968. Dictionaire etymologique de la langue grecque (Paris: Klincksieck).

Charpentier, Jarl, 1924. “Beiträge zur indoiranischen Wortkunde”, Le Monde Oriental 18: 1-45.

Dehghan, Keyvan, 1982. Der Awesta-Text Srö$ Ya$t (Yasna 57) (München: Kitzinger).

Duchesne-Guillemin, Jacques, 1936. Les composes de l’Avesta (Paris: Droz).

Fraenkel, Ernst, 1923. “Zur balto-slavischen Grammatik”, Zeitschrift für vergleichende Sprachforschung
51: 243-258.

Frisk, Hjalmar, 1952—1970. Griechisches etymologisches Wörterbuch 1—2 (Heidelberg: Winter).

Geldner, Karl Friedrich, 1907. Der Rigveda in Auswahl. Erster Teil: Glossar (Stuttgart: Kohlhammer).

Hoffmann, Karl, 1976. Aufsätze zur Indoiranistik 2 (Wiesbaden: Reichert).

Horn, Paul, 1898. *Neupersische Schriftsprache”, Grundriß der Iranischen Philologie 1/2:1--200 (Straß-
burg: Trübner).

Lentz, Wolfgang, 1926. “Die nordiranischen Elemente in der neupersischen Literatursprache bei Fir-
dosi", Zeitschrift für Indologie und Iranistik 4: 251-316.

Mayrhofer, Manfred, 1953--1980. Kurzgefaßtes etymologisches Wörterbuch des Altindischen 1-3 (Hei-
delberg: Winter).

362

Mayrhofer, Manfred, 1978. Supplement zur Sammlung der altpersischen Inschriften. (Wien: Österreichi-
sche Akademie der Wissenschaften).

Maythofer, Manfred, 1979. Iranisches Namenbuch 1 (Wien: Österreichische Akademie der Wissenschaf-
ten).

Meillet, Antoine, 1916. “À proposde (hjuta¥ma en Vieux Perse", Mémoires de la Société Linguistique
19: 348—349.

Neisser, Walter, 1930. Zum Wörterbuch des Rgveda 2 (Leipzig: Brockhaus).

Oldenberg, Hermann, 1912. Rgveda. Textkritische und exegetische Noten 2 (Berlin: Weidmann).

Pisani, Vittore, 1954. “Avest. kamaraöa-”, Zeitschrift für vergleichende Sprachforschung 71: 125-126.

Schulze, Wilhelm, 1933. Kleine Schriften (Göttingen: Vandenhoeck und Ruprecht).

Schwyzer, Eduard, 1929. "Zum SroX YaXt (y.57)”, Zeitschrift für Indologie und Iranistik 7: 97-113.

Steve, M.-J., 1975. “Inscriptions des Achéménides à Suse”, Studia Iranica 4: 7-26.

Szemerenyi, Oswald, 1977. “Indo-European Kinship” Varia 1977: Acta Iranica 16: 1—240 (Leiden:
Brill).

Turner, Ralph L., 1966. A Comparative Dictionary of the Indo-Aryan Languages (London: Oxford Uni-
versity Press).

Wackernagel, Jacob und Albert Debrunner, 1896-1957. Altindische Grammatik 1-3 (Göttingen:
Vandenhoeck und Ruprecht).

Werba, Chlodwig, 1980. “Zu einigen offenen Fragen der achaemenidischen Onomastik”, Anzeiger der
Österreichischen Akademie der Wissenschaften. Philologisch-historische Klasse (1979) 116: 13-25.

Wiist, Walther, 1966. Altpersische Studien (München: Kitzinger).

Altpersisch m-n-u-vi-i-$ = manauvis

Rüdiger Schmitt (Universität des Saarlandes)

“Eine Etymologie, die nichts weiter als
eine linguistische Möglichkeit ist, gleicht
einem Kartenhause, denn die nächste
linguistische Möglichkeit, die bekannt
wird, bläst sie um.”

(Friedrich Bechtel, Lexilogus zu Homer,
1914, S. V)

Zu den historisch bedeutsamsten altpersischen Königsinschriften der Achaimeniden gehört
zweifellos der berühmte *Fürstenspiegel' des Dareios, die Inschrift DNb. Die Zahl der mit
diesem Text verknüpften vielfältigen Probleme hat durch den überraschenden Neufund ei-
ner nahezu textgleichen Xerxes-Inschrift? , XPl, allerdings eher zu- als abgenommen. Nichts
Neues ergab sich daraus für das in diesem Text(paar) aufscheinende Hapax legomenon
m-n-u-vi-i-$, das in dem kurzen Satz aus der Selbstcharakteristik des Königs DNb 13 vor-
kommt? :

.. -ni-y mmis :ami-iy] . . . = naiy manauvi$ am]iy],

zu deutsch etwa: ‘Ich bin nicht leidenschaftlich/heißblütig/rachsüchtig/jähzornig/unbe-
herrscht’*. Im großen und ganzen ist dabei durchaus klar, was mit diesem Adjektiv gemeint
sein muß. Die genaue Bedeutungsnuance läßt sich allerdings höchstens von Morphologie
und Etymologie her feststellen, ergibt sich doch auch aus den Parallelversionen kein ganz
eindeutiger Hinweis: Die elamische Entsprechung ist |-iz-za-ma-in-da, das von Hinz (1969:
56b und 61b) zu [te}-iz-za-ma-in-da ergänzt wird und Wiedergabe von altiran. *faiavant-
o.ä. ‘mit Schärfe versehen’ (o.ä.) sein soll” ; doch verbleibt dies völlig im Unverbindlichen.
Demgegenüber steht im Babylonischen jedenfalls leidlich klares ul mam-ma Sa i-ga-a-ga
ana-ku, etwa ‘Ich bin nicht einer, der [scil.: leicht] zornig wird/in Wut gerät’ (Hinz 1969:
47a und R. Borger apud Hinz 1969: 57a).

Es gilt also zu fragen, wie die Form altpers. manauvi$ — die einzige Ambiguität in der
Graphie betrifft die Quantität des letzten Vokals — genau zu interpretieren ist und ob man
nicht über das Fazit von Brandenstein-Mayrhofer (1964: 131) “Die Bildung bleibt unklar”
hinauszukommen vermag. Dabei ist zunächst zu betonen, daß, da zweifellos ein Singular-
Nominativ vorliegt, die Form als -i-stämmig (Typus Fravarti$) oder als -i-stimmig aufgefaBt
werden muß. Für letzteren Fall wäre, argumenti causa, zu rechnen (1) mit suffixalem -i-
und Flexion entsprechend dem vedischen vrki-Typus (im Altpersischen unbezeugt),
(2) mit suffixalem -i- und Flexion entsprechend dem vedischen devi-Typus, jedoch moder-

364

nisierender Umbildung des Nominativs -i-y /-i/ zu -i-Y /-13/ innerhalb des Altpersischen (Ty-
pus Uvärazmi$ neben Uvárazmiy)! und (3) mit wurzelhaftem -i- etwa in einem Wurzelno-
men, das das Hinterglied eines Kompositums bildet.

Mit der Konstatierung dieser Fakten ist bereits gesagt, daß die etymologische Verknüpfung
von altpers. manauvi$ mit ved. manasvin- ‘sinnvoll, verständig, klug’, die zuerst Kent
(1939: 170) vorgeschlagen hat*, zum Scheitern verurteilt ist. Zwar könnte man die Umbil-
dung des Nominativs (ved. manasvi =) altpers. *manauvi zu manauviS in Analogie zu dem
entsprechenden Vorgang bei den im Singular-Nominativ gleich endenden Ländernamen des
devi-Typus im Altpersischen (vgl. oben) — Kents Operieren “with added -s” (Kent 1939:
170; Kent 1950: 65a § 187) für "manasvi/manauvis ist seltsam widersprüchlich — nach den
“Regeln der Kunst’ allenfalls noch plausibel machen. Aber dann wären immer noch die min-
destens ebenso gewichtigen Bedenken aus dem Weg zu räumen, daß (1) die (eher negative)
Bedeutung von altpers. manauvi$ keinesfalls mit der (eher positiven) des angeblichen vedi-
schen Pendants manasvin- zusammenpaßt und daß (2) der angenommene Stamm altpers.
*manauvin- der einzige Beleg für ein solches Suffix -vin- im Altpersischen wäre (vgl. Kent
1950: 52b $ 155.V), ja — schlimmer noch — daß überhaupt “außerindische Entsprechun-
gen zu -vin- ... nicht sicher nachgewiesen” sind und die “nahe Beziehung von -vin- zu
-vant- ... und -in- ... den Gedanken nahellegt), daß -vin- eine nach -in- erfolgte Umbil-
dung von -vant- ist””, mithin also eine indoarische Neuerung vorliegt, für die ein iranisches
Gegenstück gar nicht zu erwarten ist.

Mit dem einzigen weiteren in der Fachliteratur diskutierten Erklärungsvorschlag zu diesem
Wort, dem von Herzfeld (1938: 242f.), steht es jedoch nicht besser. Herzfeld hatte an ein
-i-Adjektiv zum Präsensstamm *manau- gedacht, wie er im Iranischen in, wie er schreibt
(Herzfeld 1938: 242), “manaudri-, fem. zu *manaUtar- ‘mentor, mahner”” — ge-
meint ist damit altavest. manaodri, fem. ‘gemahnend, erinnernd an etw.’ — vorliegt. Den
entscheidenden morphologischen Einwand gegen ein solches Verständnis der Form, daß es
nämlich derartige -i-stämmige Adjektive zu Präsensstämmen (und speziell solchen auf *-au-)
sonst nicht gebe, hat bereits Kent (1939: 170) erhoben!®. Dies läßt sich für das weitaus
reichere Belegmaterial des Altindoarischen jetzt auch an Hand von Wackernagel-Debrunner
(1954: 291—307 88 185—193) leicht überprüfen.

Da die bisherigen Versuche nicht zu einer überzeugenden Interpretation von altpers.
manauvis geführt haben — so schon Brandenstein-Mayrhofer (1964: 131) —, mag es gebo-
ten erscheinen, im Rahmen der oben abgesteckten Grenzen einen anderen Weg zu beschrei-
ten: Ich verstehe diese Form als manauvi$ < iran. *manah-vi-$ und als regelgemäße iranische
Entsprechung! eines unbezeugten Kompositums ved. *manas-vi- mit dem Wurzelnomen
der Wurzel vz- ‘sein Augenmerk richten auf, sich jmdm./etw. zuwenden, trachten nach
usw.’'? als Hinterglied. Da ved. mänas- neben der geläufigeren auch gar nicht so selten die
— wie das formal identische Gegenstück griech. uevos ausweist — ursprünglichere!? Bedeu-
tung ‘unwiderstehlicher Drang, zügelloser Trieb, Verlangen, Leidenschaft’ besitzt, ergibt
sich für dieses Kompositum in Analogie etwa zu ved. deva-vi ‘den Göttern zugewandt’ ein
Bedeutungsansatz ‘dem manas-, dem zügellosen Trieb, der Leidenschaft zugewandt/zuge-
tan’, also “leidenschaftlich, heißblütig, draufgängerisch, ungestüm’. So verstanden, hat es
mit altpers. manauviS durchaus seine Richtigkeit, im Formal-Morphologischen wie im Se-
mantischen; als einziger ‘Schönheitsfehler’ wäre höchstens zu bemängeln, daß eine Junktur
der beiden Kompositionselemente ar. *vi- + *manas- (als direktes oder indirektes Objekt)
bislang nicht aufgezeigt werden kann.

365

Anmerkungen

Vgl. zu dieser Charakterisierung vornehmlich Mayrhofer 1979.

Hierzu vel. zuletzt ausführlich, auch mit detailliertem Textvergleich zwischen DNb und XPI, Mayr-
hofer 1978: 21-25.

Darnach ergänzt ist XPI 14f. n-i-y : m-[n-u-vi-i-$ : a]-h-mi-i-y

= naiy ma[nauvi$ a)hmiy

mit der ‘Variante’ ahmiy statt amiy.

Aneinandergereiht sind hier nur die Übersetzungsvorschläge von Brandenstein-Mayrhofer 1964: 131;
Hinz 1969: 57b; Hinz 1975: 233; Hinz 1976: 232.

Vgl. hierzu Hinz 1969: 61b; Hinz 1975: 233 (s.v. *taifaxvanta-), aber auch die kritischen Bemerkun-
gen von Schmitt 1969: 57f.

Zu diesem Fragenkreis vgl. ausführlich Mayrhofer 1980: 134f. — Es gibt ja durchaus einige Rand-
gruppen von -i-Stämmen, die keine Motionsfeminina darstellen.

Vgl. Anmerkung 6.

Vgl. weiter Kent 1945: 47a; Kent 1950: 202a; ebenso Brandenstein 1958: 119.

So Wackernagel-Debrunner (1954: 919 8 733e), wo auf altpers. manauvis als strittige Entsprechung
verwiesen ist; zur Fragwürdigkeit dieser Gleichsetzung vgl. auch (anders als in Brandenstein 1958:
119) Mayrhofer 1963: 573£.; Brandenstein-Mayrhofer 1964: 131; Kuiper 1965: 301. — Dagegen faßt
K. Hoffmann apud Rix 1981: 117 ved. manasvin- als Umbildung von *manasva- < idg. *menes-uo-
und somit als Entsprechung von ital. *meneruo/ä-, der Grundlage des Namens der Göttin latein.
Menerva/Minerva (vgl. ausführlich Rix 1981: 111--122).

Kent (1950: 202a) schreibt lapidar: “Hz. ApI [d.h.: Herzfeld 1938] 242 otherwise, wrongly".

Für das Phonologische genügt ein Verweis auf die Gleichung ved. Särasvati- = altpers. Harauvati-
*Arachosien'.

Vgl. die eingehende Behandlung dieser Wurzel vi- (Präsens veri) bei Schmid 1968.

Vgl. etwa Schmitt (1967: 103£. u.6.); Nowicki (1976: 89f., 179 u.ö.).

Bibliographie

Brandenstein, W., 1958. Antiguo Persa. Gramatica. Inscripciones. Con un Lexico etimolögico por M.
Mayrhofer (Madrid: C.S.1.C.). |

Brandenstein, W./M. Mayrhofer, 1964. Handbuch des Altpersischen (Wiesbaden: Harrassowitz).

Herzfeld, E., 1938. Altpersische Inschriften (Berlin: Reimer).

Hinz, W., 1969. Altiranische Funde und Forschungen (Berlin: de Gruyter).

Hinz, W., 1975. Altiranisches Sprachgut der Nebenüberlieferungen (= GOF IIL3) (Wiesbaden: Harrasso-
witz).

Hinz, W., 1976. Darius und die Perser. Eine Kulturgeschichte der Achämeniden {1} (Baden-Baden:
Holle).

Kent, R.G., 1939. “The Nak$-i Rustam Inscriptions of Darius", Language 15: 160—177.

Kent, R.G., 1945. “Old Persian Texts: VI. Darius’ Naq$-i-Rustam B Inscription”, JNES 4: 39-52.

Kent, R.G., 1950. Old Persian. Grammar, texts, lexicon (= AOS 33), 2nd edition 1953 (New Haven,
Conn.: American Oriental Society).

Kuiper, F.B.J., 1965. “Review of Brandenstein-Mayrhofer 1964”, IIJ 8: 298-308.

Mayrhofer, M., 1963. Kurzgefaßtes etymologisches Wörterbuch des Altindischen, Band II (Heidelberg:
Winter).

Mayrhofer, M., 1978. Supplement zur Sammlung der altpersischen Inschriften (= SbOAW 338) (Wien:
Verlag der Österreichischen Akademie der Wissenschaften).

Mayrhofer, M., 1979. “Archaismus und Innovation. Eine grammatische Unterscheidung in Keilschrift-
Paralleltexten aus achämenidischer Zeit”, AO 47: 96--99.

Mayrhofer, M., 1980. “Zu iranischen Reflexen des rrki-Typus”, in: Recherches de linguistique. Hom-
mages d Maurice Leroy (Bruxelles: Éditions de l’Université de Bruxelles), 130 -152.

Nowicki, H., 1976. Die neutralen s-Stámme im indo-iranischen Zweig des Indogermanischen (Disserta-
tion Würzburg).

366

Rix, H., 1981. “Rapporti onomastici fra il panteon etrusco e quello romano”, in: Gli Etruschi e Roma.
Atti dell'incontro di studio in onore di Massimo Pallottino (Roma: Bretschneider), 104-126.

Schmid, W.P., 1968. “Die Wurzel vi- im Rgveda”, in: Mélanges d'indianisme à la mémoire de Louis
Renou (Paris: Boccard), 613-624.

Schmitt, Rüdiger, 1967. Dichtung und Dichtersprache in indogermanischer Zeit (Wiesbaden: Harrasso-
witz).

Schmitt, Rüdiger, 1969. "Rezension von Hinz 1969”, Kratylos 14: 54-59.

Wackernagel, J./A. Debrunner, 1954. Altindische Grammatik, Band II,2: Die Nominalsuffixe (Göttin-
gen: Vandenhoeck & Ruprecht).

Further evidence in support of Brugmann’s law

Andrew L. Sihler (University of Wisconsin)

In 1876, in the same number of Curtius’ Studien that contains his analysis of the Proto-
Indo-European nasalis sonans, Karl Brugmann printed his formulation of the sound law
generally known as Brugmann’s Law (367ff., 380-381): PIE *o alternating with *e be-
comes Indo-Iranian Zo in open syllables, but not in absolute auslaut. The examples and
categories of examples adduced have become classic: the full-grades of agent-nouns (e.g.
Skt. dataram ‘giver’ acc. sg.) vs. analogous forms in kinship terms (e.g. Skt. pitdram ‘father’
acc. sg.); o-grade o-stem nouns (e.g. Skt. bhard- ‘load, burden’, Gk. phoros ‘tribute’) next
to e-grade forms (e.g. Skt. jdnas- ‘type’, Gk. génos); root nouns like Skt. padam ‘foot’ acc.
sg. (cf. Gk. póda); o-grade forms like Skt. janu- ‘knee’ and dën. wood, Dee (cf. Gk.
gónu, dóru), next to e-grade specimens like mddhu- ‘sweet’ (cf. Gk. méthu ‘wine’); perfect
tense formations like jajna ‘he begot’ (cf. Gk. gégone) next to jajántha with its closed
syllable. As Brugmann himself noted blandly some 20 years later (1897: 138—139), the
suggestion had met with a mixed reception. Some accepted it; rather more registered
doubts, more or less on general principles; but Hübschmann and Wackernagel had sturdily
defended it, while Collitz and Schmidt had objected to it at length (see Brugmann 1897
for literature, but Hermann Collitz BB 2. 1878: 291—304). The basis for the rejection
was ostensibly the number of forms left unaccounted-for by the new sound law (there
are many, though their number shrinks dramatically if one accepts Jerzy Kurytowicz's
suggestion (1927) that laryngeals were consonants which created closed syllables in set
roots). Whatever the ostensible reason, the tone of the objections suggests that Collitz and
Schmidt found something offensive or ridiculous about the proposed law.

Probably when Kurytowicz 1927 appeared, the general view was that Brugmann's Law
leaked quite a bit, but that it held water quite well enough for a sound law. With Kuryło-
wicz’s wedding of the laryngeal theory to Brugmann’s Law, a very large number of ex-
ceptions to the law disappeared. Thus, Skt. sdkhay- ‘friend’ easily conformed to Lat.
socius and OE secg via an etymon *sokH,oy- (where the laryngeal accounts for both the
short vowel reflex in Sanskrit and the aspirated consonant), and Skt. gard- ‘swallowing’
next to Gk. borôs ‘devouring’ can be accounted for by a set root *gerH- ‘swallow’ and its
o-grade o-stem *gWorHos.

Problems remained, of course. Laryngeals shed no light on exceptions to the law like Skt.
ghand- ‘destroying’, since there is no independent evidence from Sanskrit that the root
han- had anything but anit shape. (The never-say-die or algebraic laryngealists would not
hesitate, bien entendu, to posit an otherwise unverifiable *gWhonHo- on the strength of the
short root vowel in Sanskrit.) Additionally, and perhaps most exasperatingly, this new in-

368

terpretation actually created a problem where none had existed before. This is actually
typical of new discoveries, rather than a cause for distrust per se. The excitement con-
cerned the singular of the perfect active. Brugmann and his followers held that Skt. cakdra
was the regular reflex of PIE *kWe-kWor-e, but were put to shifts to explain the Ist. sg.
cakdra, presumably < *kWe-kWor-a. Kurytowicz (1927: 103) elegantly showed that reinter-
preting the marker of the Ist sg. as *-H,¢, as suggested by the Hittite ending -(h /hi, provid-
ed the simplest of explanations for the short vowel in Sanskrit: it was in a closed syllable.
The drawback to this account is that set roots, like *genH- ‘beget’, go exactly like anit
ones: Ist sg. jajdna 3rd sg. jajdna, but the root syllable in the stem form *ge-gonH- would
have been closed in both persons, yielding jajána for both.

Surprisingly (I think), these facts led Kurylowicz to a complete volte-face as of his 1956
monograph, which rejects the validity of Brugmann's Law altogether and seeks a quite dif-
ferent explanation for the lengthenings found in Indo-Iranian. He seems to argue that
o-grades in e.g. Greek corresponding to long vowels in Sanskrit stem from the same ‘mor-
phological’ forces in language, but that the latter does not simply reflect PIE o-grades. Part
of the mechanism is something called an ‘additive proportion’, which in effect imagines
that to the Indo-Iranian Sprachgefiihl a vrddhi was the guna of a guna.’ The relevance of
this is that derived forms are more elaborately marked than primary or non-derived ones;
hence, by a somewhat involved line of reasoning, a derived form like the agentive suffix
would have full grade -tär- in Sanskrit, while a non-derived formation like the kinship terms
have the simple guna. Analogously, the morphological vitality of the words for ‘tree’ and
‘knee’, which accounts for their Gk. -o- and Skt. -a-, stems from their relationship to zero-
grades, compared to the frozen (and therefore amorphophonemic) forms like *medhu- >
Gk. méthu Skt. mddhu-. As to the perfects, his account (Kurytowicz 1956: 337) is unex-
pected: originally, Indo-Iranian would have had short vowel in both first and third singular,
such that one and the same form would have had a ‘fonction primaire’ (3rd singular) and a
‘fonction secondaire’ (1st singular). This was followed by a ‘renouvellement formel’ where-
by the third singular enjoyed an ‘introduction morphologique’ of a long vowel, in accord
with its primary function. What is unexpected about this is that, as a rule, Kurytowicz
regarded secondary/derived functions as receiving more elaborate marking than primary
ones, such that if his theories had any predictive power at all, it is surely the first singular
that should have been the focus of a ‘renouvellement formel”. However, it has persuaded
a number of astute scholars, such as Fredrik Otto Lindeman (1970: $ 36, n. 23).

In any case, the unanswered questions raised by this ‘morphological’ theory are numerous
and puzzling. Hans Hock (1974: esp. § 2.10.3) has discussed n-stem inflection (and much
else of relevance) in the Rigveda in meticulous detail, and makes a number of observations
suggesting a certain arbitrariness in the explanations offered by Kurytowicz. It is easy to
enlarge the list considerably. Kurytowicz deals with the vrddhi of Skt. svdsaram ‘sister’ acc.
sg. by saying (1956: 62) that its long vowel (and the o-grade in the cognate languages) is an
argument in favor of considering it to be a motivated form, compared, presumably, to the
underived forms in *-ser- (unattested). If it is true that forms like *dru- and *drew- account
for PIE *doru- ‘tree’ and (independently) Skt. daru-, one wonders why “tri- ‘three’ uni-
versally shows non-derived vocalism (PIE *trey-, Skt. tray-). Indeed, there are a good many
classes of root and suffix in Proto-Indo-European and Sanskrit which are transparently
‘derived/secondary’ which show e-grade and short vowels respectively: the root-vowels of
agent nouns and s-stems, not to mention such questions as why perfects were ‘formes

369

fondées’ but not imperfects or desideratives or denominatives. Additionally, why would a

transparently derivative element like the s-stem affix show e-grade and short vowels in the

oblique cases, except for *H,ewsos- ‘dawn’ (*eH,usos-?) with its opaque morphology but
o-grade (> Skt.-asam etc.) throughout?

There are two specific problems with the u-stems. We have briefly mentioned that, ac-
cording to Kurylowicz, the real explanation for why Gk. dóru gónu and Skt. daru janu are
different from the more numerous type seen in Gk. méthu Skt. madhu- is that the former
are derived/secondary, which is proved (I gather) by the existence of zero grades *dru-
*gnu- etc. If the argument is valid, the attested forms of ‘movable property, livestock’,
PIE *peku-, Skt. pasu-/pasü-, Av. pasu-, Lat. pecus, etc., are quite unexpected, since zero
grade forms are plainly attested in Skt. ksumant- and puruksu- both ‘wealthy’ (Bloomfield
1909) and as -fSu- in Avestan compounds. The attested forms are accounted for much
better by Brugmann’s Law than by Kurytowicz. A second case is less straightforward: Gk.
gnáthos and gnathmös both ‘jaw’ demonstrate the existence of zero grade forms of *genu-
‘jaw’, and, if we assume something like the etymon of the Greek forms in Proto-Indo-
Iranian (the obscurity of the -na- notwithstanding), we have an explanation for the aspi-
rated consonant uniquely attested in Skt. hdnu-: if *Znadha- (or the like) were misinter-
preted as an expression of a morphophonemic *Zhnadha-, inherited *Zanu- would be
vulnerable to remaking as *Zhanu-, thus Skt. hanu-. (To cite a clear parallel, expected Eng.
Slean lent lent < OE lénan was reinterpreted as lend etc., even in the teeth of the cognate
noun loan). If these facts are relevant, Kurytowicz would predict Gk. $gönus Skt. Shanu-.
Brugmann’s Law predicts génus and hanu-, which is closer to the mark.

It remains to defend Brugmann’s Law (aided by the laryngeal theory) against the charges
of inadequacy brought by Kurytowicz in regard to perfects like jajana ‘he begot’, and the
two o-grade types, the Proto-Indo-European iterative-causative and o-grade o-stem nouns;
according to Kuryłowicz (1956), whether the latter two types are built to set or anit roots
has no bearing on whether Sanskrit shows -2- or w- in the root vowel. This is not true, as we
shall see; but first we will discuss the sef perfects.

There is an obvious analogical avenue for the rise of nonetymological forms like 3rd sg.
jajäna, namely the proportion jagäma : jajdna : : jagäma : X; the difference between the
original 3rd sg. forms jagáma and jajána was inexplicable in Sanskrit terms, and jagäma is
vastly better differentiated. The only wonder is that the distinction failed to spread to
dadarsa ‘I, he saw’. Indeed, the obviousness and reasonableness of the analogical source for
jajana etc. is so great, that I can only assume that it is the failure of forms like $dadarsa to
appear that compromises the analogical explanation for jajdna. I suggest that the salient
trait of jagama was syllable weight rather than vowel length pure and simple. In that case, a
light but functionally ambiguous form like jajána aligned itself with the functionally un-
ambiguous jagdma, hence analogical jajana; if the functionally ambiguous dadársa aligned
itself with 3rd sg. jagama, jajana, etc., then the innovation of a formally distinct first sin-
gular would have required inventing a corresponding light form of the root. One can dream
up such — dropping the semivowel, metathesizing it, resorting to samprasarana (§dad/sa, a
sort of ‘subtractive proportion’, in the spirit of Kurytowicz) — but the fact that none of
them actually occurred carries no weight.

In Sihler 1980 I indicated that the reflexes of o-grade causative/frequentatives in Sanskrit,
so far from providing any evidence against the validity of Brugmann’s Law, as taught by

379

Kurytowicz, confirm it emphatically. Of course, a proper test requires that we confine
ourselves to roots (a) of certain Proto-Indo-European etymology and (b) which are defi-
nitely and unambiguously set or anit. Within these secure categories there are very few
forms that fail to conform to the predictions of Brugmann’s Law, pace Kurytowicz. Curi-
ously, on its face, the data show many fewer exceptions to the expected short vowel in set
roots than exceptions to expected vrddhi in anit roots (and in the later stages of the lan-
guage, there is a steady growth of vrddhi causatives to set roots, an innovation that is
hardly to be wondered at). This pattern is true also of o-grade o-stem nominals.

The following chart sorts causatives in the Rigveda which show both guna and vrddhi of
the root into Arnold's (1905) five chronological periods of the Rigveda (starting with the
earliest, they are Archaic, Strophic, Normal, Cretic, and Popular):

AS N CP

anit: gam- ‘go’ gamaya- — — — 1 -
gamaya- — — — — 2
sru- ‘hear sravaya- 2 3 1 1 -
Sravaya- - 11 - =
Sr- ‘flow’ saraya- — — — — |
saraya- - 1 - - =
set: jr- ‘waste’ jaraya- — 3 1 - 1
jaraya- — — ~ 1 -

Totals:
In accord with Brugmann's Law: 2 6 2 2 2
Contrary to Brugmann’s Law: — 2 1 22

The meagreness of the totals doesn’t disguise two facts: in aggregate the expected forms are
both earlier attested and more frequent than the unexpected ones.

Then, there are a few roots that show only one causative form, but that form is ‘wrong in
terms of Brugmann’s Law:

AS N C P
anit: ksi- ‘possess’ ksaya- — — 1 - -
nam- ‘bow’ namaya- ~ 1 1 - —
hr- ‘take’ haraya- — — 1 - -

set: am- ‘damage’ amaya- — — = —
svas- ‘blow’ svasaya- — — — — |I

The data are even more meagre here; but it is typical that the wonky set forms are later
than the anit ones. Otherwise in the Rigveda, not counting the roots from the first table
that have both guna and vrddhi forms, there are in all periods of the text 16 anit roots with
a total of 175 attestations, and 11 se/ roots with a total of 174 attestations.? It is obvious
that, so far from affording no support for Brugmann’s Law, the most reliable Rigvedic
evidence supports it much more reliably than it supports, say, the Law of Palatals.

With regard to deverbative nouns and adjectives, we are a good deal less secure about the
ablaut grade of any given formation than in causatives and perfects. For instance, there are
two types of derivatives in -zna-. Those like bhdrana- n. ‘burden’ and vacand- ‘speaking,
eloquent’ are probably allied to the e-grade Germanic infinitive seen in Go. bairan ‘to carry”.

371

Those like vacana- n. ‘a causing to recite’ and sravana- ‘causing to flow’ are often causative
in meaning, and probably continue o-grades of the root like the causative verbs. But the net
result is that one has no way of knowing for certain whether a given form is e-grade or
o-grade, and accordingly is or isn’t in accord with Brugmann’s Law.

A similar uncertainty surrounds forms presumed to be o-grade o-stem nominals. Generally
speaking, securely sef roots with vrddhi in a-stems are quite rare in the Rigveda, but they
do occur; contrariwise, anit roots without vrddhi are quite frequent in such formations in
all periods of the text, often paralleling the forms seeming to point to o-grade in accord
with Brugmann's Law: thus subhága- and subhagd- both mean *well-endowed, wealthy’, and
both are commonplace in all parts of the Rigveda. Even so, Kurylowicz (1956: 329ff. and
1977: 163—179) exaggerates the inconsistencies and misrepresents the lateness of the
vrddhi forms. Before citing the facts relating to securely anit vs. set formations, | would
like to make one general and one particular observation.

In Sanskrit, compounds are mostly made with the first element in what amounts to a stem-
form; but not infrequently the first element is in a case form: pustim-bhara- ‘bringing pros-
perity’, vajam-bhara- ‘carrying off the prize’. There are two peculiarities about such forma-
tions. (1) Every single form of such a character in the Rigveda has guna of the root in the
head of the compound, whether the root is set or anit; and (2) among anit roots there are
a total of 32 forms but they occur to only five roots, and the only securely ser root seen in
such compounds is dr- ‘pierce’, which occurs eleven times. The latter suggests that a case-
marked first element was not a realistic possibility for just any root, and the former leads
one to wonder if an o-grade formation is reflected here at all.

Presumably the first question is: are these compounds ancient compared to the stem-form
type, or are they secondary within Sanskrit? Brugmann (1906: $ 53) assumed that com-
pounds were ab origine reduced predicates, and so ab origine had the same case in the first
element that they would have had in a finite predicate. He granted that the type with stem-
form as first element is at least of Proto-Indo-European date, since they are so widely at-
tested. Indeed, since case-marked forms are actually sparingly attested, to say the least, we
gather that Brugmann's chief reasons are glottogonic, and an exactly contrary glottogonic
argument is often relied upon to account for the stem-form of compounds, namely, that
compounding itself is so ancient that it predates case-marking in syntax. Both arguments
are weak and aprioristic. In the attested history of a language like English, it can be seen
that compounds of the type painstaking, thanksgiving, and groundskeeper, showing a
morphologically marked first element, are very recent (the oldest of these three is first
seen in 1533).

One possibility is that the uniformly attested guna in the Sanskrit forms continues an
e-grade, and that they are late. This seems daring (if not incoherent), but the reasoning goes
as follows: e-grade derivatives with the verbal element governing the following form are
attested in Gk. pheré-nikos ‘carrying off victory’, a type seen mainly in names and corre-
sponding probably to Skt. trasd-dasyu- ‘name of a prince’ (‘tremble-foe’ or the like). Now,
in English the compounds that most closely approach the force of syntagms are the rela-
tively new formations in which the verb comes first: pinchpenny, killjoy, cutpurse, etc.,
that is, in the same order as the typical finite syntagm: pinches pennies, kills joy, etc. My
suggestion is that Sanskrit compounds of the type pustim-bhara- started life as representa-
tives of the e-grade frasd-dasyu- type, the rearrangement to the typical sequence of the

372

Sanskrit verb and its object including the adjustment of the ‘object’ of the ‘verb’ in the
compound to an appropriate case (not always accusative, parenthetically).

Second, and more particularly, I would like to exclude bhdga- m. ‘gracious lord’ from my
evaluation of reflexes of Brugmann’s Law. The word is specifically and conspicuously as-
sociated with deities and kings in the earliest Rigveda. The simplex and its derivatives,
especially subhaga- ‘well-endowed’ and its derivative saubhaga- n. ‘prosperity’ are very fre-
quent in the Rigveda. The conventional etymology of these forms takes them from bhaj-
‘divide, share, apportion’ thus: *‘apportioner’ > ‘generous’; the use of the word as a divine
epithet is an obvious piece of word-magic. But for one thing all sorts of divine beings are
characterized as bhdga- in the Rigveda, including the goddess Usas and, of all things, the
horses of the Asvinau. Secondly, the semantic development proposed is inside-out: notions
of ‘generous’ derive from notions of divinity and nobility, not the other way around, as in
the word generous itself < Lat. generosus ‘well-born, of good family’. Again, Eng. divine
means both ‘pertaining to deities’ and ‘uniquely wonderful, special, gratifying’, but the
latter meanings come from the former, not vice-versa. Accordingly, if bhaj- meant ‘divide’
from the beginning, it is an extremely unlikely source for bhdga- ‘lord’.

In the following tabulations of suspected o-grade o-stems occurring in the Rigveda, forms
of dubious semantics have been discarded, e.g. pama- du. ‘twins’ as if from yam- ‘reach’.
Elsewhere, forms that are too likely to be e-grades have been winnowed out (e.g. sard-
‘flowing, liquid’, without doubt is from sr- ‘flow’; but e-grade o-stems are at hand in e.g.
Lat. serum ‘whey, watery part’). Typical of the data is the set of words from i- ‘go’: gyd-
‘arrival’ (one time); triudaya- ‘threefold approach [to the altar]’ (one time) and udayd-
‘eruption’ (one time). This would be accounted two forms in rrddhi for a count of two,
and one form in guna for a count of one. From r- ‘move’ we have only ard- ‘distance’
(50 times), viz., one form in vrddhi for a count of 50.

At this rate, for the Rigveda we have the following picture of probable o-grade o-stems:

anit roots: formsind: 62 formsina: 39
count: 264 count: 278
set roots: formsina: 12 formsind: 53
count: I5 count: 634

As promised, the data for sef roots show an extreme preponderance of forms (both sort
and count) for the guna of root predicted by Brugmann’s Law. In the case of the anit
roots, the facts are both more complex and ambiguous, ambiguous because of the ever-
present possibility that an e-grade etymon underlies the form. Indeed, the three most fre-
quent forms in the ser roots have a distorting effect on the statistical picture: compounds
in -bhaga- (101 times), ksaya- m. “a dwelling’ (58 times), and bhara- ‘spoils, prize’ (43 times)
account for 73% of the total tally of anit forms in guna. The etymology of bhaj- ‘share? is
unclear, but there can be no serious possibility of a laryngeal root-extension confusing the
picture in the cases of bhr- ‘carry’ and ksi- ‘dwell’. However, since it has become clear that
Brugmann’s Law does account for the attested facts of Sanskrit in a generally straightfor-
ward way (apart from obvious leveling), I am strongly inclined to see bhära- and ksdya- as
reflexes of e-grade formations; the root accent hints at the same interpretation.

Overail, it seems to me that the foregoing data, when joined to the findings of Hock (1974)
and the general coincidence of isolated PIE o-grades with Skt. a as pointed out in 1876 by

373

Brugmann, provide stronger evidence for Brugmann's Law than that which can be amassed
in Sanskrit for the Law of Palatals. (Indeed, in view of the rather low predictive power of
the Law of Palatals, one can only marvel that Kurytowicz could have overlooked the pos-
sibility of ascribing, say, the most regular reflex of that law, the palatalization of redupli-
cated syllables in Indo-Iranian, to a purely secondary ‘morphologization’ proper to a
‘forme fondée’.) It is my hope that this contribution to the pertinent data will secure the
validity of Brugmann’s Law past reasonable question.

Notes
l This is not the place to discuss my profound doubts about the validity of Kurytowicz’s ‘additive
proportion’ in both theory and fact, but I will here point out that the best case that can be made for
some such structure operating in the history of the language is the growth of what might be called
absolute vrddhi as a characteristic of the first syllable of certain derived words. The general outlines
of the growth of this derivational mechanism are clear, and they remind one strongly of the sorts of
interactions divined by Kurytowicz in connection with his ‘additive proportion’, except that it shows
no tendency to confine itself to ablauting vowels or open syllables, as is seen in siubhaga- n. ‘pros-
perity’ (subhaga- ‘prosperous’) and därbhyä- ‘a descendent of Darbha’.

This count is somewhat smaller than the numbers given in Sihler 1980, because I have since decided
that mad- ‘be exhilarated’ and kam- ‘love’ are too doubtful in etymology to be included; and also
that vr- ‘cover’ and vr- ‘choose’, which can pretty confidently be traced to anit and set etyma, re-
spectively, have become too confused morphophonemically to be used as evidence.

References

Arnold, E. Vernon, 1905. Vedic metre in its historical development (Cambridge: the University Press).

Bloomfield, Maurice, 1909. “On some disguised forms of Sanskrit pasu ‘cattle’”, Indogermanische For-
schungen 25: 185—199.

Brugmann, Karl, 1876. "Zur Geschichte der stammabstufenden Declinationen. Erste Abhandlung: Die
Nomina auf -ar- und -tar-", Studien zur griechischen und lateinischen Grammatik 9: 361—406.

Brugmann, Karl, 1897. Grundriß der vergleichenden Grammatik der indogermanischen Sprachen, vol. 1,
part 1, 2nd ed, (Straßburg: Trübner).

Brugmann, Karl, 1906. Grundriß der vergleichenden Grammatik der indogermanischen Sprachen, vol. 2,
part 1 (Straßburg: Trübner).

Hock, Hans Henrich, 1974. “On the Indo-Iranian accusative plural of consonant stems”, JAOS 94: 73—
95.

Kuryłowicz, Jerzy, 1927. “ə indo-européen et 4 hittite”, in: Symbolae grammaticae in honorem Ioannis
Rozwadowski, vol. 1 (Kraków: Universytet Jagielloński).

Kurytowicz, Jerzy, 1956. “L’apophonie en indo-européen”, Prace jezykoznawcze, vol. 9 (Wrocław:
Polska Akademia Nauk).

Kuryłowicz, Jerzy, 1977. “Problèmes de linguistique indo-européenne”, Prace jeęzykoznawcze, vol. 90
(Wroctaw: Polska Akademia Nauk).

Lindemann, Fredrik Otto, 1970. Einführung in die Laryngaltheorie (= Sammlung Göschen vol. 1247—
1247a) (Berlin: de Gruyter).

Sihler, Andrew L., 1980. "Review of Kurytowicz 1977 (q.v.)", Language 56: 110—113.

Greek gaXtos, Latin *balan, Old High German bal (?) ‘marked by a blaze’:
a horse fanciers’ multilingual symposium

Otto Springer (University of Pennsylvania)

Linguists and classical philologists* — the man whom we are honoring today can pride him-
self on being both — have long been intrigued by the more than casual interest in languages
and linguistics displayed by the Byzantine historian Procopius of Caesarea who flourished
under the East Roman emperor Justinian (527—65) and whose magnum opus are the eight
books on the history of his master’s wars.’ In addition to the Greek of his day and to Latin
which still served as the official language in all parts of the empire (so that even Theodoric
the Great condescended to indicate his imperial approval by signing Latin LEGI with the
help of a golden stencil), Procopius, born in Palestine, may have considered Syrian his
native tongue; he refers to Phoenician and Armenian more than once; repeatedly and, as we
shall see, with special interest, he alludes to the language of his archenemies, the Persians;
and in one instance, we have reasons to believe, he brings in Gothic — in Procopius’ words,
the yawn ForB km Xeyouévn, which he says the East and West Goths as well as the Vandals
have in common.

To give just a few examples. It so happened that during the military siege of the town of
Petra in Kolchis an important role was played by a substance whose native Iranian de-
signation, nafth (Modern Persian naft, also nift according to F. Steingass, A Persian-English
Dictionary p. 1416), appears as vág0a in Procopius' Greek text, the very form that has
been adopted by almost all languages of the world today ; moreover, the author was rather
specific about adding the Greek equivalent as he introduced the new and foreign term:
ayyeta 5€ Gewv TE Kal dapaNrov Eutinodpevoi Kai papudkov dreo Mido pév vapdav
kaAodaw, EAAnmves de Mnöelae EXaıwov “And they had filled pots with sulphur and bitumen
and the substance which the Persians call ‘naphtha’ and the Greeks ‘Medea’s oil’.” (VIII,
11,36)

In a similar ‘bilingual’ vein Procopius, the native of Caesarea which is el-Kaisérije today,
treats the reader to the Persian adoption of a loanword from Latin: Katoap, orco yàp
Tov 'Pcouawov Baotréa Kadovot Ilépoas (II, 21, 9) “Caesar, for thus the Persians name the
king of the Romans” — in contrast to the term flactAeUc used for the Greek-Byzantine
emperor. Incidentally, the student of Old High German is reminded of that early attempt
by the author of the medieval German Annolied, about 1100 A. D. (ed. by M. Roediger,
1895: 121), at etymologizing the German counterpart of the same word:

duo santin si den edelin Cesarem,

dannin noch hiude kuninge heizzint keisere. (vv. 271f.)

376
Il

In the light of these and many similar instances which have been collected and critically
reviewed by E. Schwyzer (1914: 303—27) — Procopius received none-too laudatory marks
from his belated professorial reader —, a third ‘bilingual’ passage from Procopius’ Gothic
War may deserve our renewed attention. It tells of that dramatic moment in the war be-
tween the East Goths and the Romans with Belisarius near the Milvian Bridge (537 A.D.),
when Belisarius relinquished his army command and started fighting in the front ranks.
What might have caused his undoing was his horse which, while unusually experienced in
battle, stood out because “this whole body was dark grey (or brown?), except that his face
from the top of his head to the nostrils was the purest white." At that point Procopius
adds the linguistic remark: “Such a horse the Greeks call ‘phaliòs’ and the barbarians
‘balan’.’’ In the author's Greek: ... ds ön ÖAov uev TO Ge yawos NV, TO HETWTOD dé
àmar Ek kepanc uéxpr Es bas SAeUk ds HAALOTG. TOUTOV EdAnves pév yaridv, BapBapot
è Badap Kadovat. (V, 18, 6)

F. Wrede, the first to draw our attention to this passage (1891: 58 Anm. 2), at first blush
suspected a form of Germanic wala, walha ‘foreign’ in the “barbarian” word. However, he
quite obviously neglected the context which required a semantic equivalent of Greek
pads ‘marked by a blaze’. Thus it was left to E. Schroder in the same year (1891: 237ff.)
to come out with a very challenging analysis of the word. He started with the Greek term
vac (and its partly synonymous qaAóc, qaAapóc? ,) for which the meanings of "shining,
white’ as well as ‘dappled’ and most specifically ‘marked (on the forehead) by a blaze’ were
well attested not only by early Greek authors, such as Euripides, Aristophanes, or Calli-
machos and their commentators but also by ancient lexicographers like Hesychius (ed. M.
Schmidt 1862: 4, 228f.): yadapdv: \euKdv; padapds: yadtos, padax pdc, EUKOHÉTUIMOS,
AeukÓc, kai qaAÀéov; qaXóv- [Aaunpóv]; eaA[A]óc - AevKds, or Herodianus (ed. A. Lentz
1867: I, 123. 16): yaluôs ò Aevrouerwrros.”

And since the stem form of the Greek adjectives, yad-, presupposes a PIE base with *bA-,
reflected as *b- in Germanic *bal-, Schröder was inclined to see in its “barbarian” counter-
part the inflected (strong or weak acc. sg.) form of a Gothic nominal derivative, such as
*balan (from *bals). Its cognates in West Germanic are the -k- (< PIE -g-, -g-) derivatives
Old High German belicha, -0, Old Low German beliko (see Althochdeutsches Wôrterbuch
1952—: 1, 871), Middle High German belche, New High German (dialectal) Belch{e), Bel-
chen ‘bald coot’ (“with a slate grey body but a shiny-white blaze from the forehead to the
root of the bill”). Even more striking a parallel is the recurrence of Belche as the name of
Dietleib’s horse in the medieval German epic Biterolf (see Deutsches Heldenbuch, 1866—
71: I w. 2275, 2687, 11973). In fact, Schröder did not hesitate to rediscover the word also
in various topographical designations like the mountain name Belchen, probably from a
bald (or snow-covered?) feature of the terrain.

No doubt, in this particular instance, the linguistically inclined historian of ancient times
succeeded in lining up not only two semantic equivalents in two different languages but at
the same time two etymological cognates — needless to say, without being able to ration-
alize them as such. Indeed, the root from which both of them derive, PIE *bhel[2)-, also
with lengthening grade *bhel-: *bhol- and zero grade *bhol-, is well represented in the Indo-
European family of languages.*

377

. There is Old Indic bhala- ‘shine; forehead’ (< *bhelo-), going back to the same root as Old
Indic bhati 'shines? , Avest. vohvavant- ‘provided with good light or splendor’(?) (< *vohu-
wa-vant-?)6 ; Latin fulica (dialectal for *folica < *bholikä) ‘bald coot”” ; Alban. balé (< *bal-
ja) ‘horse with a blaze on his forehead’* ; Armen. bal ‘paleness, pallor’? ; Old Church Slavic
belu (< *bhelo-) ‘white’, Russ. belyi, Pol. biaty ‘same’'° ; Lithuanian bälas besides the more
usual baltas (<*bolstos) and balnas (< *bolanos) ‘white’'', Latvian bäls (: bals) ‘pale,
pallid’, baits ‘white’? , Old Prussian ballo ‘forehead’ (Elbing Vocab. 77)? ; other Germanic
formations in -/- with lengthening grade are most probably Old English bae/ n., Old Norse
bál (< *bhélo-) ‘fire, flame, pyre’’* ; rather controversial (Erbwort or loan?) and much in
need of clarification are most of the Celtic cognates, such as Welsh bal adj. ‘with a blaze’ or
Middle Bret. baill ‘same’'*; also the ever-repeated hypothetical Gallic *balio- has been
viewed with some scepticism!é. — Mention must further be made of the well-known name
of Achilles’ immortal horse, Badtos (with retracted accent, Ilias Book 16, 149; 19, 400)!”
as well as of the adjective and appellative Badtds (< *baliuds?), whose initial stops (instead
of spirants, cf. Schwyzer 1953: 1, 291) suggest a non-Greek, perhaps Phrygian, Thracian,
or Illyrian loanword. As is evident from some lexicographical observations in ancient times,
there was little doubt even then as to the semantic concurrence of yadtds and Barus:
bards... f) qaMóc Ttc. Gv, è Aeukouérconoc Kat ò Xeukóc (Etymologicon Magnum 186,
27ff.) and ro yalur 56 kat Baduov: Aéyovow &ni row €xóvrcw 7t XevkÓv ÈV TG) uercoon
(Scholia in Theocritum vetera 8, 26.)!?

To be sure, since the initial *bh- of the PIE root "*bAel(2)- emerges as a spirant only in
Greek and Latin but every where else in Europe as a stop, E. Schróder's assumption that the
“barbarian” word balan cited by Procopius was of Gothic origin seems — in the absence of
unequivocal phonological criteria — to be largely supported by circumstantial evidence;
Procopius does use the expression ‘barbarian’ often interchangeably with ‘Goth(ic)’. Still,
on closer scrutiny circumstances in this instance do turn out to be rather cogent, for one
additional reason: the striking blaze with which Belisar’s horse happens to be marked
would by itself hardly bring the great Roman soldier into mortal danger, had it not been,
as Procopius tells us, for those Roman mercenaries who sympathized with the Goths and
therefore had deserted to their ranks. Of course, they knew Belisar as well as the horse he
was riding, so they tried to have the Goths concentrate their attack on Belisar by shouting
at them to aim “at the balan horse”; but their Hetzruf only made sense if they could hope
that the Goths would understand which horse (and with it, which rider) they were eager to
point out to them. |

In fact, we have reason to believe that some of those Roman soldiers were no less familiar
with that “barbarian” term than the Goths and the learned historiographer who reports it.
Procopius himself certainly had no monopoly on it. For during his time the word happens
to occur in Latin garb and context in the heading of a poem written by the Christian
author Magnus Felix Ennodius, born and raised in Arles on the Rhone river, from 513—21
Bishop of Pavia, who composed a number of — painfully unwitty — epigrams, one of them
entitled De equo badio et balane (see Ennodi Opera. 1885: 246 Carm. No. 2, 136): ‘About
the Chestnut-Colored Horse with the Blaze on His Forehead’ (the variant badeo of the

378

manuscript reflects Vulgar Latin). We also have knowledge, through one lonely attestation
from the year 1354, of a medieval French adjective balani ‘wei&gefleckt’; it occurs in a
metrical line: on roncin bron, la teste balanie (listed in Godefroy 1880: 1, 560). Not
enough, Tiktin in his Rumänisch-Deutsche Wôrterbuch (1903: 1, 146f.) carries quite an
extensive article on the Rumanian adj. and subst. bälan (bäla’ik) with such meanings as
‘blond, light’; ‘white’ (also of the hair of domestic quadrupeds); moreover cal balan ‘Schim-
mel’ as well as Balan by itself used as a proper name for oxen, cows, dogs and sheep. Small
wonder that this Germanic loanword for ‘marked with a blaze (on the forehead)’, which
during the broadening contacts between the Germanic tribes and the ancient world had
evidently found its way into Vulgar Latin everywhere, from the Balkans across Italy and
into southern France, was not lost on a rather sophisticated intellectual — and “front-line
war correspondent” — Procopius of Caesarea (See Suchier, 1894: 186ff. and Loewe, 1906:
299ff.).

We may be surprised to find the designation of a horse with that particular marking of his
hair borrowed from a Germanic dialect by speakers almost through the entire ancient
world. And yet, as we go through the files of early Greek medical treatment of horse
diseases, that is, as we leaf through the corpus of Greek hippiatrics and try to analyze the
designations of horses by their color of hair (Corpus Hippiatricorum Graecorum 1927: 2,
312 No. 201.), we discover in a treatise which E. Schwyzer (1929: 66, 93-99) — perhaps
somewhat too boldly — has in part dated back to the fifth century A.D., that out of a
dozen names at least four are of undoubtedly Germanic provenience: ó yaAßas ‘der Falbe’,
cf. Old High German falo (< Germanic *falwa-) ‘hellbraun, fahl, falb’; 6 BAayxas ‘the white
horse’ (German ‘Schimmel’), cf. Old High German blank ‘weiß, glänzend’, Old English
blanca m. ‘white horse’; 6 yeBas (with -8- = [b] or [v], cf. Greek AerBua for Livia, Schwy-
zer 1953: 1, 208) ‘the grey- (or dun-)colored horse’, clearly from a (non-attested) Gothic
*grews (<*Germanic *grewaz, Old High German grá[w]o, Old English graew), the first
unequivocal phonological criterion favoring Gothic as the point of origin because in con-
trast to West and North Germanic only Gothic would have a long -e- of very closed quality
(it is frequently written ei, the Gothic grapheme for long i).

What is left is — last but not least — 6 Badas: kaAoc kac ebrovoc ‘the horse with the blaze
on his forehead, beautiful and vigorous’, the “barbarian” designation which we had tended
to trace to Gothic from the very beginning. Thus the word BaAas, far from being restricted
to Belisar's horse, must have made its way, like ypas and the rest, from Gothic into
Greek, too. As a matter of fact, Procopius does not say that Belisar's horse ‘bore the name
of Balan" in the barbarian idiom and that of “Phalios” in Greek but rather makes the
general linguistic observation that “such a horse the Greeks designate as ‘phalios’ and the
barbarians as ‘balan’.” Concerning the declensional pattern, E. Schwyzer (1929: 1, 97f.)
and before him E. Schröder (1891: 238) as well as M. Schónfeld (1911: XXIVf.) rightly
suggested that the frequently used Gothic form in -an served as a point of departure for
switching the borrowed word into the Greek -ac/-av declension; Latin may have gone a
step further and, as in the case of "extensions" like Attilanem, Totilanem used by Jordanes
and others, may have carried the -v- through the system, hence Ennodius’ Lat. abl. sg.
balane (see above) and the form with -n- or -n, and even -ni in Romanic.

379
IV

In view of the impressively dense and wide distribution of forms derived from the PIE root
*bhel(a)- etc. over the entire realm of Indo-European and of the rather convincing case that
can be made for the use of the simple stem form, Germanic *bal-, in Gothic to designate
something like ‘marked by a blaze’, it is surprising to find not a trace of it in North Ger-
manic; also within West Germanic only the before-mentioned residual derivations, Old High
German belicha, -o and Old Low German beliko with their -k- (<PIE -g-, -g-) suffixes,
testify indirectly to the one-time productivity of *bal- within that limited subarea of West
Germanic. There is, however, in addition possibly one case of direct evidence — though far
from being uncontested, as we shall see —, a solitary Old High German gloss listed in Stein-
meyer-Sievers 1879-1922: 1, 259, 31. It happens to be recorded in identical form as pal in
two rather old and closely related manuscripts, namely K, otherwise designated as Sg 911
of the Old High German Abrogans Glossary (formerly referred to as Keronisches Glossar),
from the 8th century and written in the Alemannic dialect, and Ra, the Reichenau manu-
script of the Abrogans Glossary, otherwise designated as Cod. Carlsruhe Aug. CXI, from
the 10th century and likewise in Alemannic, although both go back to a (lost) Bavarian
original. Its Latin lemma is spelled flaus instead of *flauus '(hell) glánzend, (weiß)leuch-
tend', as we have reasons to rewrite it, — not only in accordance with the Latin glosses:
thoca, flauus uel uestis and Flauus, genus uestis (Corpus Glossariorum Latinorum 1894: 5,
612, 35 and 5, 599, 64 respectively) which must form the lexicological background, but
also in accordance with the fact that the faulty writing of Latin flavus with one 4 must
have been so frequent that Probus Donatius Servius in the Appendix of his De Arte Gram-
matica warns: “flavus, not flaus" (see Grammatici latini 1855—80: 4, 198, 5 and Thesaurus
linguae latinae 1900: 6, 1, 887). And as to the spelling of the Old High German interpreta-
mentum pal, the initial p- for Germanic b- is to be expected in Upper German manuscripts
of the 8th and 10th centuries (see Braune 1975: § 136 Anm. 1).

However, the entire gloss which in Steinmeyer-Sievers loc. cit. reads synoptically across the
page: flaus: pal (K) — pal (Ra) and flaus: stigma [from Greek o7tyya ‘puncture, character-
istic mark”, orryun once for ‘spot on a bird’s plumage”] (from the parallel passage in R, the
Viennese manuscript, Cod. Vindob. 162 of the so-called Samanunga) and which is preceded
by an equally problematic line: Toga: stoz (K) — stoz (Ra) — Toga: lahhan, has remained
so obscure that again and again the entire complex has been suspected of textual corrup-
tions. Hardly worth mentioning is the notion of Schlutter (1912/13: 14, 147) through
“emending” flaus to *fluus or *flaus (with a deleted -a-) to come up with a supposedly
earlier form of New High German Flaus 'Überrock" which, in fact, is a relatively recent loan
from Middle Low German (cf. Kluge 1967: 203). Also Schenk (1912: 46) with his “ex-
planation” stôz st.f. ‘Toga, Obergewand’ obviously has nothing at all to contribute. Much
more serious is the attempt by J. Splett, author of a very thorough “running commentary”
on the Abrogans (1976: 390f.), through a combination of gloss 1,259,31 with 3,152,40
and 3,191,61 to identify bal with what elsewhere is recorded as Old High German bello or
bal meaning the ‘ball to play with’, to shift the entire section and what precedes it to the
subject matter of toys. This shift, however, forces him to give the word stéz in the same
passage the highly improbable meaning of ‘ball, spinning top” — a tour de force about
which Splett himself has his doubts. Needless to say. if the solution should at all lie in that
direction, Old High German pal, i.e. bal, would have nothing to contribute to the topic of
this paper.

380

However, one more possibility must be touched on briefly in closing: as is evident from a
very short article in Du Cange (1840--50: 3, 320), as early as the 17th century the word
flavus was “ingeniously” restored to clavus, meaning ʻa strip (or patch) of cloth (of a dif-
ferent color) on the Roman tunic' (see Thesaurus linguae latinae 1900—: 3, 1330; Corpus
glossariorum latinorum: Thesaurus glossarum emendatarum 1899—1901: 6, 456; 7, 353;
Splett 1976: 390). This emendation seems eminently plausible because in a dictionary of
"synonyms" (“Sachgruppen” would be a more adequate term), such as the Abrogans,
Latin clavus after a fashion goes well with foga and the obviously “sartorial” context of
the whole gloss — the illustrations cited in Thesaurus linguae latinae loc. cit. again and
again combine clavus with toga, vestis, and tunica. The emendation also would eliminate
the adjectival masculine ending us which clashes with the gender of every noun in the
passage. It might even suggest a possible rationale for the occurrence of Old High German
stöz there because the very first of the more technical senses of the word Stof listed in
Deutsches Wórterbuch 1854—1954: 10, 3, 473 is “Zeugstreifen’ (“in der älteren Zeit oft
kostümlich besonders ausgestaltet"), i.e. ‘patch, piece’, German ‘Flecken, Fleck(en)’; or in
Upper German dialect dictionaries like the Schweizerisches Idioticon 1881--: 11, 1585ff.
‘Saumbesatz’; see also Adelung 1774—80: 4, 789. Strangely enough, the basic meaning of
Latin clavus, too, seems to refer to that part of a (woven) fabric where it borders on (‘an-
stößt’) and contrasts with another (cf. Pauly-Wissowa 1894—1963: 4, 4ff.)

However, what about Old High German pal/bal? The answer will largely depend on how
precise and reliable our knowledge will eventually be of the relative chronology of głos-
satorial work on the Abrogans; also in this regard Splett holds out hope for further sub-
stantive progress. Yet, in order not to prejudice the case, it may be wise under present
circumstances to assume that the glossator who inserted the Old High German translation
pal/bal in the sense of ‘(hell)glänzend, (weiß)leuchtend’ already had the Latin color ad-
jective flavus before him, no longer *clavus, for which pal/bal as an Old High German
interpretamentum would make no sense at all. Thus we are inclined — not without resigna-
tion -- to accept the solution also presented by the Althochdeutsches Wörterbuch (1952-:
1, 786), a solution which for the time being is based on the actual juxtaposition of the
adjectives *flavus and *bal, that is, flaus and pal, and which allows us to point to an exact
Old High German counterpart of the Gothic word shouted by Roman mercenaries at the
Goths to bring down Belisar, their own former commander, and reported in a Greek
narrative, almost 1500 years ago, by the Byzantine historian, Procopius of Caesarea.

Notes

* The research on which the following paper is based was done in connection with the work toward an
Etymologicai Dictionary of Old High German carried on with the generous support of the National
Endowment for the Humanities.

Procopii Caesarensis Opera Omnia (rec. J. Haury) I-III (Leipzig 1905—13). All references are to this
edition. — For an English translation see H.B. Dewing. I--VH (Loeb Classical Library) London, New
York, Cambridge, Mass. 1914-40.

2 See Frisk 1960-72: 2, 988f.; Boisacq 1950: 1013f.; Schulze 1892: 463f.

See also F. Jeschonnek, De nominibus quae Graeci pecudibus domesticis indiderunt. (Diss.) Regens-
burg, 1885.

381

^ See Walde-Pokorny 1927-32: 2, 175f.; Pokorny 1959-69: 118ff.; Persson 1891; 20. 109; Persson

1910-12: 28f.; 117f., Anm. 2; 569 Anm. 1; Lidén 1897: 76f. and fn. 5.

See Mayrhofer 1953-80: 2, 493; 496f. (doubtful).

See Bartholomae 1904: 1435.

See Walde-Hofmann 1938-56: 1, 559f.

See Meyer 1891: 24 (cf. also Sitzungsber. d. Akad. d. Wiss. in Wien 130 [1897], 69) and Jokl 1915:

35.

? See Scheftelowitz 1905: 37.

10 See Trautmann 1923: 29ff.; Sadnik-Aitzetmüller 1955: 10: Sadnik-Aitzetmüller 1963-: No. 129;
Berneker 1908-13: 1,55; Vasmer 1950-70: 1, 73.

See Fraenkel 1955-65: 1, 32.

See Mühlenbach-Endzelin 1923-32: 1, 272.

See Trautmann 1910: 310.

See Falk und Torp 1909: 267; Holthausen 1963: 15; Jöhannesson 1953: 625f.; Vries 1962: 23;
Holthausen 1948: 10, Feist 1939: 77f.

15 See Dictionary of Welsh 1950—: 250; Loth 1899: 344f.; Loth 1900: 396f.; Thomas 1900: 452;
Pedersen 1925: 91ff.; Jackson 1967: $ 1174.

See Stokes 1894: 164; Holder 1896-1913: 1, 336, 370ff., 386f.; Dottin 1920: 230; Meyer-Lübke

1935: No. 906; Wartburg 1928: 202 (badius), 217 (balla); modified by Bertoldi 1931: 291 fn. 2;
Whatmough 1949: 289.

See Lexikon des frühgriechischen Epos 1951-: 2, 23 (“wahrscheinlich ‘Schecke’”).

See Solmsen 1897: 75 (favoring Phrygian origin of BaAıos; Detschew 1957: 41 (listing also BdAac,
Baans, Balius used as proper names). Cf. also Krahe 1955: 1, 115; Brandenstein 1950—52: 76; Pokor-
ny 1959—69: 118ff.

D A nn

li
12
13
14

17
18

References

Adelung, J.C., 1774--80. Versuch eines vollständigen grammatisch-kritischen Wörterbuchs der hoch-
deutschen Mundart. Leipzig.

Althochdeutsches Wörterbuch, 1952-. Bearbeitet und hrsgg. von E Karg-Gasterstädt, Th. Frings et al.
Berlin.

Bartholomae, Chr., 1904. (Reprint 1961). Altiranisches Wörterbuch (2. unveränd. Aufl.). Berlin.

Benecke, G.F., W. Müller, and F. Zarncke, 1854-61. Mittelhochdeutsches Wörterbuch. Leipzig.

Berneker, E., 1908-13. Slavisches etymologisches Wörterbuch. Heidelberg.

Bertoldi, V., 1931. Revue Celtique 48: 291 fn. 2.

Boisacq, E., 1950. Dictionnaire étymologique de la langue grecque (4. Aufl.). Heidelberg.

Brandenstein, W., 1950—52. Die Sprache 2, 76.

Braune, W. and Eggers, H., 1975. Althochdeutsche Grammatik (13. Aufl.). Tübingen.

Corpus Glossariorum Latinorum, 1894. Ed. G. Goetz. Leipzig.

Corpus Hippiatricorum Graecorum, 1927. Ed. E. Oder & C. Hoppe. Leipzig.

Detschew, D., 1957. Die thrakischen Sprachreste. (Österr. Akad. d. Wiss., Philos.-hist. Kl.: Schriften d.
Balkan-Kommission. Ling. Abt. 14). Wien.

Deutsches Heldenbuch, 1866—71. Hrsg. von O. Janicke. Berlin.

Deutsches Wörterbuch, 1854-1954. Hrsg. von J. und W. Grimm. Leipzig.

Dictionary of the Welsh Language, 1950--. Caerdydd.

Dottin, G., 1920. La Langue gauloise. Paris.

Du Cange 1840—50. Glossarium mediae et infimae aetatis. Paris.

Ennodi Opera, 1885. (rec. F. Vogel) (Monumenta Germaniae Historica: Auctorum Antiquissimorum T.
7). Hannover.

Falk, H. and A. Torp, 1909. Wortschatz der germanischen Spracheinheit (4. Aufl.). Göttingen.

Falk, H. and A. Torp, 1960. Norwegisch-dänisches etymologisches Wörterbuch. Oslo.

Feist, S., 1939. Vergleichendes Wörterbuch der gotischen Sprache (3. Aufl.). Leiden.

Fracnkel, E., 1955—65. Litauisches etymologisches Wörterbuch. Heidelberg.

Frisk, H., 1960-72. Griechisches etymologisches Wörterbuch. Heidelberg.

382

Godefroy, F., 1880. (Reprint 1969). Dictionnaire de l’ancienne langue française. Paris.

Graff, E.G., 1834-46. Althochdeutscher Sprachschatz. Berlin.

Grammatici latini, 1855 —80. Ed. H. Keil. Leipzig.

Herodiani Technici Reliquiae, 1867. Ed. A. Lentz, Leipzig.

Hesychii Alexandrini Lexicon, 1862. Ed. M. Schmidt, Jena.

Holder, A. Th., 1896—1913. Altceltischer Sprachschatz. Leipzig.

Holthausen, F., 1948. Vergleichendes Wörterbuch des Altwestnordischen. Göttingen.

Holthausen, F., 1963. Altenglisches etymologisches Wörterbuch (2. Aufl.). Heidelberg.

Jóhannesson, A., 1953. Isländisches etymologisches Wörterbuch. Bern.

Jokl, N., 1915. Indogermanische Forschungen (Anzeiger) 35, 35.

Kluge, F., 1975. Etymologisches Wörterbuch der deutschen Sprache (21. Aufl. bearbeitet von W.
Mitzka). Berlin.

Krahe, H., 1955. Die Sprache der Illyrier. Heidelberg.

Lexer, M., 1872-78. Mittelhochdeutsches Handwörterbuch. Leipzig.

Lexikon des frühgriechischen Epos, 1955 —. Göttingen.

Lidén, E., 1897. Studien zur altindischen und vergleichenden Sprachgeschichte. (Skrifter utg. af K.
Humanistiska vetenskaps-samfundet i Uppsala, 6, 1). Uppsala.

Loewe, R., 1906. Altgermanische Elemente der Balkansprachen. Zeitschrift f. vgl. Sprachforschung 39,
299-302.

Loth, J., 1899. Remarques sur le Wortschatz der Keltischen Spracheinheit de M. Whitley Stokes. Revue
celtique 20, 344—5.

Loth, J., 1900. Archiv f. keltische Lexikographie 1, 396—7.

Mayrhofer, M., 1953—80. Kurzgefaßtes etymologisches Wörterbuch des Altindischen. Heidelberg.

Meyer, G., 1891. Etymologisches Wörterbuch der albanischen Sprache. Straßburg.

Meyer-Lübke, W., 1935. Romanisches etymologisches Wörterbuch (3. Aufl.). Heidelberg.

Mühlenbach-Endzelin, 1923—32. Lettisch-deutsches Wörterbuch. Riga.

Pauly-Wissowa, 1894-1963. Realenzyklopädie der classischen Altertumswissenschaft. Stuttgart.

Pedersen, H., 1925. Litteris 2, 91—3.

Persson, P., 1891. Studien zur Wurzelerweiterung und Wurzelvariation. (Uppsala universitets arsskrift,
4). Uppsala.

Persson, P., 1910—12. Beiträge zur indogermanischen Wortforschung. (Skrifter utgivna af Kgl. Humanis-
tiska vetenskaps-samfundet i Uppsala, 10). Uppsala.

Pokorny, J., 1959—69. Indogermanisches etymologisches Wörterbuch. Bern und München.

Prellwitz, W., 1897. Studien zur idg. Etymologie und Wortbildung. Bezzenbergers Beiträge 22, 80f.

Prellwitz, W., 1905. Etymologisches Wörterbuch der griechischen Sprache (2. Aufl.). Göttingen.

Procopii Caesarensis Opera Omnia, 1905 —13. (rec. J. Haury). Leipzig.

Procopius. With an English translation by H.B. Dewing. 1914-40. London, New York, Cambridge,
Mass.

Sadnik-Aitzetmüller, 1955. Handwörterbuch zu den altkirchenslavischen Texten. Den Haag, Heidelberg.

Sadnik-Aitzetmüller, 1963—. Vergleichendes Wörterbuch der slavischen Sprachen. Wiesbaden.

Scheftelowitz, J., 1905. Zur altarmenischen Lautgeschichte. Bezzenbergers Beiträge 29, 37.

Schenck, O., 1912. Zum Wortschatz des keronischen Glossars. (Diss.) Heidelberg.

Schlutter, O.B., 1912/13. Glossographische Beiträge zur deutschen Wortgeschichte. Zeitschrift f. dt.
Wortf. 14, 137—60.

Schönfeld, M., 1911. Wörterbuch der altgermanischen Personen- und Völkernamen. Heidelberg.

Schröder, E., 1891. Belisars Ross. Zeitschrift f. deutsches Altertum 35, 237—244.

Schulze, W., 1892. Quaestiones epicae, Gütersloh.

Schulze, W., 1910. Etymologisches. Sitzungsberichte d. kgl. Preuß. Akad. d. Wissenschaften, Philol.-hist.
KI. 1910, 2. Berlin.

Schweizerisches Idioticon, 1881-. Bearbeitet von F. Staub, L. Tobler, A. Bachmann u.a. Frauenfeld.

Schwyzer, E., 1914. Die sprachlichen Interessen Prokops von Cásarea. Festgabe Hugo Blümner 303—27.
Zürich.

Schwyzer, E., 1929. Germanisches und Ungedeutetes in byzantinischen Pferdenamen. Zeitschrift f.
deutsches Altertum 66, 93-99.

Schwyzer, E., 1953. Griechische Grammatik (2. Aufl.) München.

Solmsen, F., 1897. Thrakisch-Phrygisches. Zeitschrift f. vgl. Sprachforschung 34, 72—77.

383

Splett, J., 1976. Abrogans-Studien: Kommentar zum ältesten deutschen Wörterbuch. Wiesbaden.

Starck, T. and J. C. Wells, 1972—. Althochdeutsches Glossenwórterbuch. Heidelberg.

Steinmeyer-Sievers, 1879—1922. Althochdeutsche Glossen. Berlin.

Stokes, W., 1894. Urkeltischer Sprachschatz (4. Aufl.). Góttingen.

Suchier, H., 1894. Chlotars des II. Sachsenkrieg. Zeitschrift f. roman. Phil. 18, 186--9.

Theocritum, Scholia in. . . vetera. 1914. (rec. C. Wendel). Leipzig.

Thesaurus linguae latinae. 1900—. Leipzig.

Thomas, A., 1900. Review of: V. Henry, Lexique étymologique des termes les plus usuels du breton
moderne. Romania 29, 452.

Tiktin, H., 1903—. Rumánisch-deutsches Wórterbuch (Dictionar Román-German). Bukarest.

Trautmann, R., 1910. Die altpreußischen Sprachdenkmäler. Göttingen.

Trautmann, R., 1923. Baltisch-slavisches Wörterbuch. Göttingen.

Vasmer, M., 1950—70. Russisches etymologisches Wörterbuch. Heidelberg.

Vogel, Fr., 1884. Ennodiana. Archiv f. lat. Lexikographie 1, 270.

Vogel, Fr., 1885. Balteanus. Archiv f. lat. Lexikographie 2, 477—8.

Vries, J. de, 1962. Altnordisches etymologisches Wórterbuch (2. Aufl.). Leiden.

Waide, A. und J. B. Hofmann, 1938—56. Lateinisches etymologisches Wörterbuch (3. Aufl.). Heidelberg.

Walde, A. und J. Pokorny, 1927-32. Vergleichendes Wörterbuch der indogermanischen Sprachen. Ber-
lin und Leipzig.

Wartburg, W. v., 1928--. Französisches etymologisches Wörterbuch. Berlin, Leipzig, Basel.

Whatmough, J., 1949. Review of J. Pokorny, Idg. etym. Wörterbuch, Fasc. 1: Language 25, 289.

Wrede, F., 1891. Über die Sprache der Ostgoten in Italien. (Quellen und Forschungen 68). Straßburg.

Further evidence for diachronic selection: Ved. rästi, Lat. regit etc.

Klaus Strunk (Universität München)

In two articles recently published (1981, 1982) I tried to show that in various Indo-Euro-
pean languages or even in dialects of a single language we often can find atype of related,
but not wholly corresponding units the differences among which must be explained by
generalization of one or another variant resulting from invariants in prehistoric (predia-
lectal) paradigms, sometimes in such of the parent language itself. If this view is correct,
the real existence of proto-languages like Common Aryan, Common Greek and ultimately
Proto-Indo-European somehow must be acknowledged since it is presupposed by the
inevitable assumption of earlier basic linguistic systems with certain characteristic gram-
matical subsystems (inflectional categories). There are numerous clues for such phenomena
in the historical Indo-European languages unified into either direction from former alter-
nants (invariants or allomorphs) in their proto-language. They are often well known as, for
instance, the different lexicalizations of stem-alternants in neuter nouns of an old hetero-
clitic inflection. In the Aryan subgroup of Indo-European one should remember, in this
framework, the diverging root-initial consonants in Vedic present and aorist forms of gam
‘to go’ on the one hand and in their Old Iranian counterparts on the other. The Vedic verb
has g- throughout its synchronic paradigm (pres. ind. gicchati, aor.ind.-inj. [d]gan, subj.
gdmat, opt. gamyat, causative gamayati), whereas the equivalent Avestan one offers cor-
responding forms beginning with j- (pres.ind. jasaiti, aor.ind.inj. [uz jön Y. 46.12, subj.
Jimat, opt. jamiiat, OP ajamiya DPd 19, causative jamaiieiti). This hardly can be understood
otherwise than owing to respective selection and generalization of different variants stem-
ming from a Common Aryan paradigm wherein a phonologically conditioned alternation
between forms with plosive *g- and others with palatalized *)- must have existed: e.g. pres.
ind. *gascati < *gWmsketi, aor.opt. *gamyat < *gWmiea,t, causative *gamayati <
*gWomejeti vs. aor ind.-inj. *(aJJan « *(eJgWemt, subj. *Jamat(i) X. *gWemet(i).

Many linguistic data — familiar to Indo-Europeanists — occurring with similar slight varia-
tions in several Indo-European languages may be interpreted thus. I do not aim at going on
to enumerate examples of this kind here in order to corroborate the indicated view in
general. But in return it is perhaps worth calling attention to the possibility that sometimes
the principle of diachronic selection suggested here can be helpful to solve hitherto remain-
ing problems of comparative findings in detail. The discrepancy in vowel-quantity between
Ved. rasti, rajati ‘reigns, rules’ and Lat. regit ‘directs, rules’ besides further cognates with
short root-vowel in some other Indo-European languages may serve as a specimen for such
an interpretation.

386

The Sanskrit verb is attested with forms of the athematic root-present two times in the
Rgveda. In the first passage rdsti because of its syntagmatic lexical relation to prá . . . tirate
belongs evidently to the inherited root */a/reg- meaning something like ‘to stretch, direct,
rule’: RV. I 104,4 pré pürväbhis tirate rasti sürah "in former (times? ) the (Aryan) hero
pushed forward and ruled’. The meaning of 3rd pres.inj. rat in RV. VI 12,5 seems to be
less clear. Geldner (1951:104) translated 'eilt" with question-mark added in brackets and
thought (in the footnote) of a mutilated form belonging to vraj ‘to go, wander’. Yet I
presume that the lexical sense of rdt here is partly comparable to that of rdsti in RV. I
104,4. In the hymn VI 12, devoted to Agni, within the first three padas of the fifth tristubh-
stanza the celebrated brightness (panayanti bhdso) of the god is mentioned, when he,
starting like a quick runner, goes lopping through the country; then we read in the fourth
pada: yno nd tayur dti dhänva rat ‘ike an indebted robber he rat all over waste landscapes’.
For rat a translation ‘s’étend’ (= 'extends") was proposed by Renou (1964b:133), who
moreover referred to a potential connection between verbal 3rd sg.inj. raf VI 12,5 and the
nominal syntagma ral agnis ‘ruler Agni’ in the beginning of the same hymn (RV. VI 12,1).
In any case not only RV. I 104,4 with rasti, but also RV. VI 12,5 with rat — this time in a
metaphor of spreading fire — could allude to the Aryans’ conquest pushing on, whereby
they became ‘rulers’. The meanings of ‘expansion, extent, stretch, direction’ can be com-
patible, for pertinent implications of sacral kingship see Gonda (1956:passim), Schlerath
(1960:105s.); the objections raised by Sihler (1977) are not cogent and coupled with
uneven new suggestions, see below. In the Rgveda there is for the rest no necessity to
distinguish a verbal root raj 2 ‘herrschen, gebieten’ with thematic present-stem rdja- (as e.g.
RV. I 188,1, equally said about Agni, sámiddho adyá rajasi devó .. . sahasrajit ‘inflamed
thou rule today, being a god, ..., thou, having been conqueror thousand times") and a
root raj 1 ‘glänzen, strahlen, hervorleuchten, sich auszeichnen’ comprising the two attesta-
tions of an athematic root-present-stem raj- and likewise numerous examples of a themat-
icized stem raja- (homonymous to raja- 2), as is done by Grassmann (1955:1156s.) and
others (e.g. Wackernagel-Debrunner 1954:3). An original distinction between raj 1 ‘strah-
len’ and raj 2 ‘herrschen’ in the Rgveda has seriously been doubted by scholars like Olden-
berg, Noten I 81; Geldner in his Rig-Veda-translation, passim; Renou (1964a:12; Kratylos
10:102 (1965)); Mayrhofer (1976:56). An intimate link between ‘Kénigtum’ and ‘Glanz’
in the Veda was emphasized by Schlerath (1960:105). A typologically parallel compatibil-
ity of the meanings ‘bright’ and ‘ruler, king’ was pointed out by Eilers (1967:94 with n. 4),
in Accadian “Sarrum ‘König’ = ‘glänzend’”.

If we thus can uphold in general the link between Ved. rästi, rajati ‘reigns, rules, beams’
(RV.), the root-noun rdt (rdj-) ‘king’ and some verbs outside the Vedic language like Av.
"razaiti (see below), Lat. regere etc. (Pokorny, 1959:854—857), there remain nevertheless
some problems to be discussed. At first the root-inflection of Ved. rasti seems to be old,
having been preserved within the Rgveda only in two relic-forms. A transition into the
more productive thematic inflection rajati must have taken place rather early, as is in-
dicated by Av. ‘razaiti (see below). But how can we understand the long vowel of rästi to
have arisen from a root */a/reg- with short -é-? Sihler's (1977:233—236) attempt to derive
the long vowel of Ved. rásti, ràj- ‘king’ and its Proto-Indo-European predecessor from the
laryngeal of a hitherto unknown root *rea,g- (separated from "reg-) ‘be efficacious, have
mana’, which he would see elsewhere only in Ved. #rj- (f.) ‘nourishment, power, strength’,
can scarcely be right. The idea clearly depends upon including Ved. urj- in order not to be

387

ad hoc only for Ved. rasti and the related noun rat, Lat. rèx (and derivatives), Olr. rí ‘king’.
But instead of ürj- we should expect $ i7j- from a zero grade “ra,g- (without y- !) in word-
initial position (see Wackernagel 1896:28), so that this combination must be abandoned
(other etymological suggestions for ürj- are adduced by Wackernagel 1896:261; Mayrhofer
1976:652). Now, are there other possibilities for explaining the -4- in the Aryan verb? The
length of the root-vowel in the repeatedly attested Avestan iterative-stem razaiia- (razaiieinti
Yt. 14.56 etc.) ‘to direct, stretch, arrange’ is worthless here since it may be due to Brug-
mann’s law, for which Goth. ufrakjan ‘to stretch out’ with short -a- from *-ó- and its Ger-
manic congeners like OE reccan, OHG recchen etc. are possibly comparable. More inter-
esting is the long vowel in the verbal adjective Av. raSta- (besides raStam Yt. 14.43) ‘di-
rected, arranged’, OP. rdsta- ‘straight’, Turfan Pahl. raS¢ ‘true’: it must be analogical,
presumably borrowed from a present "raz(a)- corresponding to Ved. ráj(a)-! . Old Iranian
*razati is attested in Av. virazaiti ‘rules exactly squaring with Ved. vi-rajati ‘reigns’ RV. X
189,3 etc.: Yt. 14.47 varadraynam ... yO virazaiti antara rasta rasmana ‘VereÜrayna ...,
who rules between the two lines of battle ranged’. Consequently we have to reckon with
an Indo-Iranian present-stem *raz(a)-. Its long root-vowel must be accounted for.

Generally it is held that Ved. rasti or its Aryan predecessor is a denominal derivative of the
inherited root-noun Ved. rat, raj- (cf. Lat. rex, Olr. r£), so that the long vowel would have
its source in the nominal basis (Mayrhofer 1976:56 with further references; Joachim 1978:
145). But this explanation is doubtful for two reasons. At first, denominative verbs derived
without an additional suffix from athematic noun-stems are unusual, as is rightly argued
also by Sihler (1977:231s.). Rare examples as Ved. bhisäkti, Av. bifaz- ‘heals’, often sup-
posed to have been based on bhisak (bhisdj-) ‘healing, physician' (Whitney 1889:387;
Wackernagel-Debrunner 1954:152; Joachim 1978:144) may be primary (attested in Vedic
and Avestan) and independent of the Vedic noun (compare Kuiper 1937:44—47, Mayr-
hofer 1963:503). Secondiy, of rästi, if denominative in origin from raj- ‘king’, only the
meaning ‘reigns, rules’, as in RV. I 104,4, but not ‘extends, pushes on’, as in rat RV. VI
12,5 discussed above, would be semantically conceivable. A similar objection must be
raised against Sihler's (1977) results since the meaning of his suggested root *rea,g- ‘be
efficacious, have mana’ does not fit for rat RV. VI 12,5 apart from the fact, that his pro-
posal suffers from further difficulties (see above). Thus to my mind it should be maintained
that the noun (Ved. rat, Lat. réx, Olr. ri ‘king’) and the verbs Ved. rasti, rdjati, Lat. regit
etc. all go back to the Indo-European root */a/reg-. Then, however, we cannot help seeking
for another motivation of the vowel-length in the Indo-Iranian verb and of its contrast to
short -č- in Lat. regere and some cognates in other Indo-European languages.

A solution of this problem can be offered in the framework of the ‘proterodynamic’ or
‘acrostatic’ root-presents discovered by Narten (1968), i.e. of an inflection with stem-grada-
tions unknown so far. In this athematic inflection-type an accent firmly established on the
first syllable (‘acrostatic’) was combined with an alternation between lengthened grade in
forms of the strong stem and full grade in forms of the weak one (having additionally zero
grade in gradable endings). Thus we have, as Narten showed, 3rd sg.pres.ind.act. Ved.
Stáuti vs. med. stáve, 3rd sg.pres.ind (inj.) act. Ved. tasti, Av. tast(i) vs. 3rd pl.ind.act. Ved.
taksati etc. This ‘acrostatic’ root-inflection must have existed already in the parent lan-
guage. That is proved by remnants as 3rd pl. Ved. dsate, Gk. etarai (for farai) “they sit’,
Ved. 3rd sg. säve = Gk. keiraı ‘lies’, all being Media tantum’ with full grade of the root.
Forms as Ved. 3rd pl.act. stuvanti RV. (instead of *stavati, cf. 3rd pl. taksati) besides 3rd

388

sg. stduti were remodelled after the more usual type bearing mobile accent (e.g. dógdhi vs.
duhánti). The idea of Insler (1972:60—61) according to which the plural-active-forms of
acrostatic presents would originally have had — like those of sigmatic aorists — lengthened
grade, too, is purely theoretical and scarcely correct since Insler is forced to interpret the
full grade of the really attested forms Ved. (pl.) ataksma, atasta, táksati, taksur as ana-
logically shaped after the weak stem of the participle taksat-. Elsewhere I tried to show
that the lengthened grade in all athematic inflectional categories with ‘acrostatic’ accent
(singular indicative-injunctive active of root-presents, indicative-injunctive active of sigmatic
aorists, strong cases of an acrostatic heteroclitic neuter like *jekWr/iékWn- liver’) cannot
have been engendered phonetically by the immobile accent (being causal only for the
primary full grade-forms), but must be due to prehistoric morphological rearrangements
(‘Sekundäraufstufungen’) aligned after mutatis mutandis comparable paradigms (with
alternations caused by mobile inflectional accent) in order to introduce stem-gradations
secondarily to acrostatic paradigms that previously had only forms with full grade-stems
(Strunk 1985).

Whatever may be true in this respect, the acrostatic inflection-type itself meanwhile seems
to be well founded. It was vindicated with additional arguments e.g. by Narten herself
(1968:15, n.44) for the verb ‘to eat’ with continued and generalized lengthened grade in
Lith. é[d]mi, ésti, OCS jami, jasti, OLat. es, est vs. Ved. átti, by Eichner (1973:81) for
Hittite 1st sg.pret. wékun ‘I demanded’ vs. 31d pl.pres. wekkanzi, by Beekes (1973:92) and
Schindler (1975:267) at the outset of lexicalized doublets like Gk. 700s vs. Edos, ynpas
vs. Yépas etc., by myself for an athematic prototype as the basis of Lith. bégu, bégti, OCS
béZo, béZati ‘to run’ on the one hand and of Gk. yeßopaı, péBeodat ‘to flee’ with revealing
middle voice on the other (see Strunk 1985:493—495) etc. These explanations, as far as
Baltic and Slavic forms are concerned, are preferable to Winter’s (1978:433), according to
which Lith. ésti, OCS jasti, Lith. bégti, OCS béZati would have got their lengthened vowels
by way of Winter’s tentative sound-law (Indo-European short vowels before non-aspirated
voiced stops are supposed to have become long in Baltic and Slavic). For at first there are
several exceptions contrary to Winter’s law, the most important of which is Slav. voda
‘water’ from *vod-; and secondly the expressly attested vowel-length of OLat. és, est (see
Leumann 1977:529) continuing the same inherited verb as Lith. ésti, OCS jasti cannot be
justified by Winter’s law suggested only for Baltic and Slavic: this means that competitive
explanations starting from previously acrostatic paradigms are farther-reaching than Win-
ter’s since they comprise, beyond the borders of Baltic and Slavic, comparable phenomena
in further Indo-European languages as Latin (Greek, Hittite, Indo-Iranian etc.).

With regard to such other more or less ensured occurrences of Indo-European acrostatic
root-presents the relic-forms Ved. rästi, rat (RV.) may be attributed to this type of inflec-
tion, too. They probably continue inherited forms of the 3rd sg.pres.ind. and inj.act.
*f ajreg-ti, *(ajrég-t belonging to the normal strong stem and alternating with forms of the
weak stem in the root-paradigm of */a)reg- like 3rd pl. *(2/rég-nti, ptc. *(2)rég-nt-, 3rdsg.
subj. *(2Jrég-e-ti; for full grade of the root in subj.act. of acrostatic inflections compare
Tichy (1976:79) and in particular the sigmatic aorists disclosing a characteristic opposition
within the Ved. type ind. nais- vs. subj. nesa- etc. On the other hand the weak-grade-stem
of the primitive paradigm with full grade of the root seems to have been generalized in
some related verbs of other Indo-European languages. The most representative example of
them is Lat. regere, rego, regit etc. ‘to direct, rule’ with secondary thematic vowel compar-

389

able to that of Myc. e-u-ke-to, Gk. eüxerau, grown up from an old acrostatic root-present
visible in Ved. 3rd pl. óhate, ptc. óhana-, Av. 3rd. sg. aoxte, 3rd pl. aojaite, which is ad-
duced by Narten (1968:10—12).

Probably many subsequent formations of this sort with thematic vowel were developed
by passing an intermediate stage of subjunctives like */a 1Jéugh-e-toi, *(a)rég-e-ti etc. It
should be noticed that rise of certain thematic present-stems from reassigned subjunctives,
once belonging to athematic indicatives/injunctives, often has been assumed in general and
in detail (thus e.g. Kuiper (1937:95—104); Risch (1981:702—709); Tichy (1976:79—80,
n.21); Bammesberger (1982:340—343)). Outside Latin we find correspondent formations,
which can be supposed to have been likewise derived from the weak stem (subjunctive) of
the old acrostatic paradigm “/o)rég-/ (ajrég(e)-, in the following verbs: Goth. rikan, with
considerably shifted meaning in 2nd sg. rikis translating owpevoers R. 12.20 ‘you will
heap’; MHG and NHG rechen ‘to heap’, perhaps from ‘to erect, stretch (out)’ (see Seebold
1970:373); Gk. dpéyew ‘to reach, stretch out’ with a prothetic vowel the origin of which
(23-?) is much disputed about. Olr. rigid ‘stretches out’, generally held to contain the same
stem */2/rege- as the verbs just mentioned (see Ernout-Meillet 1959:568; Frisk 1970:413;
Chantraine 1961:817), is traced back to a present-formation with suffix *-je-/-jo- by
Vendryes-Bachellery (1974:R 14). If so, the Old Irish cognate must be added to a number
of other verbs in Indo-European languages that, though being equally derived from the root
*(a)reg-, do diverge in some different and at least in part not inherited stem-formations:
e.g. Ved. rij(á)-, rjya-, irajyd- (with enigmatic initial vowel /-), Lith. ré2ti, réZiu (with com-
bination of a nasal and a *-je-/-jo- present); their word-histories need not be treated of here.

Whereas Lat. rego and its Greek and Germanic counterparts reflect the weak stem of
the prehistoric root-present with acrostatic accent, presumably via the subjunctive-stem
*/9 jrége/o-, the thematic stem with radical -d- of Ved. rajati, Av. ‘rdzaiti besides Ved. rasti
must have developed somewhat otherwise. A striking parallel is Ved. dasati ‘offers, grants,
honours’ besides dasti id.’ RV. I 127,4, that has obviously been supposed to be an acro-
static root-present, too (Narten 1968:15, n. 43; Tichy 1976:79). If Gk. Sékoua (Att.
ôéxoum) with short -e- (like Lat. rego etc.) goes back to an Indo-European subjunctive of
the primary acrostatic present, as Tichy suggested, the length of -d- in Ved. dasati just as
that in Ved. rajati, Av. ‘razaiti can have been introduced only in the Aryan subgroup.
Tichy (1976:78) argued that within the synchronic layer of the Rgveda there still seem to
be subjunctives with a stem dasd- as dasad RV. IV 2,9, where this mood-function of the
stem is indicated by the neighbouring unequivocal subjunctive-forms krnävate (in the same
relative clause) and vosat, varat (in the following principal clause). There is surely no doubt
that other occurrences of the verbal stem dasá- clearly function as indicative-forms. Even a
new subjunctive-stem with lengthened thematic vowel -a- had already been developed:
ddsàt RV. 168,6; 71,6; 933; I1 234. Consequently the language of the Rgveda apparently
disposed of a subjunctive dasat (belonging to ind. dasti) besides a more recent type dasar
(squaring with ind. dasati). That means, the pres.ind. dasati got its finally attested shape in
early Vedic or perhaps Aryan times, when the original alternation between acrostatic sg.
ind.act. *uekri (> Ved. dästi) and subj. *dekeri (cf. Gk. öekopaı) was replaced by a levelled
stem with generalized -ã- of the root as in the subjunctives däsat and däsat. In accordance
with these subjunctive-stems containing a long root-vowel the thematic present ddsati was
remodelled from an earlier dásati (cf. Gk. 6€kouat), that perhaps is still hidden in Ist pl.
opt.act. disema RV. VII 3,7a, where a short root-vowel is required by the trochaic cadence

390

of the tristubh verse. The renewal of a presumable Aryan *razti > Ved. rasti (RV.) to Ved.
rajati, Av. "rázaiti probably covered the same way. There are also occurrences of a subjunc-
tive with long root-vowel, early attested in Ved. Ist sg.act. virajani RV. X 159,6; 174,5 and
AV. I 29,6. It cannot be decided, of course, whether this form contains a subjunctive-stem
räjä- adherent to ind. rasti, or raja- squaring with the thematic ind. rdjati (compare dsäni
besides asti like vahani besides vahati). In any case, however, the Indo-Iranian present Ved.
räjati, Av. °ràzaiti besides *razti > Ved. rasti seems to have been completed, because of its
long radical -4-, not before the Aryan stage in the history of language just like the subjunc-
tive 1st sg.act. *rdzd > Ved. rdja(ni) with generalized length of the root-vowel for its part.

To sum up: Ved. rasti, rdjati, Av. ‘razaiti and subj. Ved. 1st sg. rajani reflect or are enlarged
from a PIE acrostatic present */a rest’ , subj. */a/régeti. In other words, the long root-
vowel belonging to the strong stem (singular indicative/injunctive active) of the inherited
paradigm was selected and generalized. Therefore it is not necessary to take Ved. rasti or,
as did Gonda (1956:164), rdjati for denominal derivatives of *rdz-, Ved. raj- ‘king’ with
unusual word-formation; at best the meaning and the morphological development of the
thematic verb Ved. rajati, Av. "razaiti in being rearranged from an Indo-Iranian subjunctive
may have been favoured by a vague relation to the noun *raz- ‘king’ on the basis of an
analogical alignment after a ‘paronymic” pair consisting in the predecessors of Ved. pati-
‘master, lord” and patyate (synchronically analysable as pdty-ate) ‘possesses, rules’. Neither
we are forced to assume with Sihler (1977) a hitherto unknown and isolated root *rea,g-
especially for the root-noun *rég-s and the verb Ved. rasti, rajati (for Gk. apryyw see below
n. 5). On the other hand Lat. regere, Gk. ópéyeu? and Goth. rikan with their short root-
vowel *-é- exclusively continue the weak stem of the Proto-Indo-European acrostatic
paradigm, presumably via its prehistoric subjunctive. Among these primary verbs Lat.
regere contains the inherited potential meaning ‘to direct, rule’ besides primordial ‘to
stretch”, that is better visible in the ppp. rectus, in compounds like erigere, por(ri)gere,
surgere etc. and is the only preserved one in the related Greek and Germanic verbs. The
secondary, i.e. previously iterative verb OE reccan, however, means, obviously not by
chance, like Lat. regere not only ‘to stretch, extend’, but also ‘to rule, direct, guide’.

Starting from a Proto-Indo-European acrostatic present ind.act.sg. */a/régti, subj. */a)régeti
the resultant Indo-Iranian primary verbs ultimately seem to be based on the first, the Latin,
Greek and Germanic ones on the second of these prehistoric stem-alternants.

Notes

See Bartholomae (1904:110 8 209.7, n. 11); borrowing of -d- in Av. ra§ta-, OP rdsta- from a (not
attested!) s-aorist, as suggested by Kent (1953:34, n. 2 to § 93) is highly improbable.

An Av. rdz- ‘gehen’, assumed by Bartholomae (1904:1526) exclusively for Yt. 14.47, is not required
by the text. The translation given above makes good sense and the conformity between Av. virdzaiti
and Ved. vi-rdjati ‘rules, reigns’ should not be disregarded (K. Hoffmann, by letter).

Schindler (by oral communication) compares OAv. razarö (n.) ‘statute, order’, instr. sg. rašnā with
alternation of lengthened and full grade in an acrostatic heteroclitic noun of the same root. — Is the
generalized lengthened grade of Lat. rex, regis including derivatives (regius, regalis, etc.) and of Olr.
rí. gen. ríg (see Sihler 1977:226) due to early levelling of an acrostatic alternation, *reg- -/rég- in
strong and weak cases? Whatever may have happened in this paradigm, there are parallels in other
root-nouns with uniform lengthened grade of the stem: compare Lat. ér, gen. éris, Gk. xjp ‘hedgehog’
from *gher-, Gk. Ojp, 9npds, Lith. Zvéris from *ghuer- (Lat. férus) etc. and see Schindler (1972:37).

391

^ The term ‘paronymy’ has been coined by Winter (1969:43—45), who emphasized the relevance of

this notion for understanding certain analogical processes.

If Gk. bpéyew comes from this root, Gk. apryew ‘to help’, 4onyuov ‘helper’ because of their devi-
ating initial vowel should be left aside in spite of Schulze (1933:172 n. 3), Gonda (1956:165). This
etymological starting-point of Gk. aptiyw, apnyudv is rejected by Frisk (1960:137) for semantic
reasons. Sihler (1977:236) would combine Gk. ap7yw, apwyds with *rég- ‘king’ etc. on the basis of
his suggested root “rea £-

Benveniste (1969:14) assumed an original meaning 'to align' (denoting special functions of a sacral
kingship as e.g. regere fines).

References

Bammesberger, A., 1982. “Einige e-stufige Präsentien des Urgermanischen und ihr Verhältnis zu indo-
germanischen athematischen Wurzelaoristen”, PBB 104:339—344.

Bartholomae, Ch., 1904. Altiranisches Wörterbuch (Straßburg: Trübner).

Beekes, R.S.P., 1973. “The Proterodynamic ‘Perfect’”, KZ 87:86—98.

Benveniste, É., 1969. Le vocabulaire des institutions indo-européennes, Tome 2 (Paris: Éd. de Minuit).

Eichner, H., 1973. “Die Etymologie von heth. mehur”, MSS 31:53—107.

Eilers, W., 1967. “Stern, Planet, Regenbogen. Zur Nomenklatur der orientalischen Himmelskunde”, in:
Der Orient in der Forschung. Festschrift für O. Spies zum 5. April 1966, edited by W. Hoenerbach
(Wiesbaden: Harrassowitz), 92—146.

Frisk, H., 1960. Griechisches etymologisches Wörterbuch 1 (Heidelberg: Winter).

Frisk, H., 1972. Griechisches etymologisches Wörterbuch 2 (Heidelberg: Winter).

Geldner, K.-F., 1951. Der Rig-Veda 2 (Cambridge, Mass.: Harvard UP).

Gonda, J., 1956. “Semantisches zu idg. r&g- “König” und zur Wurzel re$- “(sich aus)strecken” ”, KZ 73:
151-167.

Grassmann, H., 1955. Wörterbuch zum Rigveda (3. Auflage) (Wiesbaden: Harrassowitz).

Insier, St., 1972. *On Proterodynamic Root Present Inflection", MSS 30:55—64.

Joachim, U., 1978. Mehrfachpräsentien im Rgveda (Frankfurt/Bern/Las Vegas: Lang).

Kent, R.G., 1953. Old Persian. Grammar. Texts. Lexicon, Second edition, revised (New Haven, Con-
necticut: American Oriental Society).

Kuiper, F.B.J., 1937. Die indogermanischen Nasalpräsentia (Amsterdam: N.V. Noord-Hollandsche
Uitgeversmaatschapij).

Leumann, M., 1977. Lateinische Grammatik Band 1: Lateinische Laut- und Formenlehre (München:
Beck).

Mayrhofer, M., 1963. Kurzgefaßtes etymologisches Wörterbuch des Altindischen 2 (Heidelberg: Winter).

Mayrhofer, M., 1976. Kurzgefaßtes etymologisches Wörterbuch des Altindischen 3 (Heidelberg: Winter).

Narten, J., 1968. “Zum ‘proterodynamischen’ Wurzelprisens”’, in: Pratidanam. Indian, Iranian and Indo-
European Studies presented to F.B.J. Kuiper on his sixtieth birthday (The Hague: Mouton), 9--19.

Oldenberg, H., 1912. Rgveda, Textkritische und exegetische Noten. Siebentes bis zehntes Buch. (Berlin:
Weidmann).

Pokorny, J., 1959. Indogermanisches etymologisches Wörterbuch 1 (Bern/München: Francke).

Renou, L., 1964a. Études védiques et päninéennes XII (Paris: de Boccard).

Renou, L., 1964b. Études védiques et päninéennes XII (Paris: de Boccard).

Risch, E., 1981. “Zum Problem der thematischen Konjugation” (1965), in: E. Risch, Kleine Schriften,
edited by A. Etter and M. Looser (Berlin/New York: de Gruyter) 702—709.

Schindler, J., 1972. “L’apophonie des noms-racines indo-européens", BSL 67:31—38.

Schindler, J., 1975. “Zum Ablaut der neutralen s-Stämme des Indogermanischen”, in: Flexion und
Wortbildung, edited by H. Rix (Wiesbaden: Reichert), 259—267.

Schlerath, B., 1960. Das Kónigtum im Rig- und Atharvaveda (Wiesbaden: Harrassowitz).

Schulze, W., 1933. Kleine Schriften (Göttingen: Vandenhoeck & Ruprecht).

Seebold, E., 1970. Vergleichendes und etymologisches Wörterbuch der germanischen starken Verben
(The Hague/Paris: Mouton).

Sihler, A.L., 1977. “The etymology of PIE *rég”- ‘king’ etc.”, JIES 5:221—246.

392

Strunk, K., 1981. “Stammbaumtheorie und Selektion”, in: Logos Semantikos. Studia linguistica in
honorem E. Coseriu 1921—1981, Vol. 2, edited by H. Geckeler — H. Weydt et al. (Berlin/New York:
de Gruyter/Madrid: Gredos), 159—170.

Strunk, K., 1982. “‘Vater Himmel’ — Tradition und Wandel einer sakralsprachlichen Formel", in: Serta
Indogermanica. Festschrift für G. Neumann zum 60. Geburtstag, edited by J. Tischler (Innsbruck:
Institut für Sprachwissenschaft), 427—438.

Strunk, K., 1985. “Flexionskategorien mit akrostatischem Akzent und die sigmatischen Aoriste”, in:
Grammatische Kategorien. Funktion und Geschichte. Akten der VII. Fachtagung der Indogerma-
nischen Gesellschaft Berlin, 20.—25. Februar 1983, edited by B. Schlerath (Wiesbaden: Reichert),
490—514.

Tichy, E., 1976. “Gr. &eiéxaro und idg. *dékti, déktoi", Glotta 54: 71—84. d

Vendryes, J./E. Bachellery, 1974. Lexique étymologique de l'irlandais ancien. Lettres R S (Paris: Ed.
du C.N.R.S.).

Wackernagel, J./A. Debrunner, 1896. Altindische Grammatik I (Göttingen: Vandenhoeck & Ruprecht).

Wackernagel, J./A. Debrunner, 1954. Altindische Grammatik 11/2 (Göttingen: Vandenhoeck & Rup-
recht).

Whitney, W.D., 1889. Sanskrit Grammar Including both the Classical Language, and the older Dialects,
of Veda and Brahmana, second edition (Leipzig: Breitkopf & Härtel/London: Kegan Paul, Trench,
Trübner & Co). -

Winter, W., 1969. “Analogischer Sprachwandel und semantische Struktur”, FoL 3:29-45.

Winter, W., 1978. “The distribution of short and long vowels in stems of the type Lith. &szi : vesti :
mesti and OCS jasti : vesti : mesti in Baltic and Slavic languages”, in: Recent Developments in
Historical Phonology, edited by J. Fisiak (The Hague/Paris/New York: Mouton), 431—446.

tulyasyaprayatnam savarnam Astädhyäyi (1.1.9)

K.M. Tiwary (Patna University)

It is generally agreed upon that Panini's grammar does not undertake the phonetic analysis
and description of Sanskrit speech sounds. Rather it aims at formulating maximally brief
and general rules to describe the phonological/morphophonological processes of Sanskrit.
It is also agreed upon, however, that some of these rules presuppose phonetic analysis and
descriptions of Sanskrit speech sounds as given by the tradition from before. One such rule
is tulyasyaprayatnam savarnam (1.1.9), which constitutes the concept of homogeneity of
speech sounds and defines it in phonetic articulatory terms. Obviously, the purpose of this
rule is to give an intensional definition in terms of which the class of homogeneous speech
sounds can be formed and used to describe the morpho-phonological processes of Sanskrit
with generality and economy. Briefly, the rule may be glossed as follows: let the speech
sounds which are articulated at the same place in the supra-glottal region and by the same
type of articulatory efforts be called homogeneous speech sounds. Thus, to construct
classes of homogeneous speech sounds it is first necessary to specify the places of articula-
tion and the types of articulatory effort.

1. Places of articulation

The supraglottal region is treated as equivalent to the whole mouth, and the mouth is
held to extend from the lips at one end to just before the glottis at the other. That is, the
glottis is excluded from the supra-glottal region of the vocal tract. This region is further
sub-divided into six functional sections for articulatory purposes. These sections are (1)
kanthah (the pharynx); (2) jihvamula (the root of the tongue); (3) talu (the palate); (4)
mürdhä (the highest point of the palate); (5) dantah (the teeth); (6) osthau (the two lips).
Some of these places of articulation are selectively combined to give rise to further articu-
latory positions necessary for characterizing sounds which cannot be accommodated within
the sections already marked off. They are as follows: (1) kanthatalu (the pharynx-and-
palate); (2) kanthostham (the pharynx-and-lips) and (3) dantostham (the teeth-and-lips).
Besides, nasika (the nose) is also listed as a place of articulation.

Traditionally, the best known and widely used description of Sanskrit sounds in terms of
the places of articulation enumerated above is given below. It would become evident later
that Pänini’s rule cited in the title of this paper presupposes this system of description:

(1) akuhavisarjaniyánam kanthah, i.e. [a; k, kh, g, gh; : (visarga) and h are articulated in

the pharynx.]

394

(2) icuyasanam talu, i.e., i: c, ch, j, jh; y ands are articulated at the palate.
(3) rturasanam mürdhä, i.e.,r; t, th, d, dh; r and s are articulated at the highest point of
the palate.
(4) Itulasanam dantah, i.e., I; t, th, d, dh; l and s are articulated at the teeth.
(5) upüpadhmaniyanamosthau , i.e., u; p, ph, b, bh; : p are articulated at the lips.
(6) fiamananananam nasika , i.e.,7í; m; h; n and n are articulated in the nose.
(7) edaitoh kanthatalu, i.e., e, ai are articulated at the pharynx-and-palate.
(8) odautoh kanthostam , i.e., o, au are articulated at the pharynx-and-ips.
(9) vakarasya dantostham, i.e., v is articulated at the teeth-and-lips.
(10) jihvamülivasya jihvamüulam, i.e., jihvamülam (7 : K) is articulated at the root of the
tongue.
(11) nasikanusvarasya, i.e. anusvar (7 ©) is articulated in the nose.

It is clear that this system of phonetic description and classification in terms of the place of
articulation groups together speech sounds of functionally different phonological classes,
viz., consonants and vowels. For instance, a, i, u, r, l are grouped together with velar,
palatal, bilabial, cacuminal and dental consonants, respectively. Even phonetically distin-
guishable sounds are not kept apart in this system of classification by place of articulation.
For instance, r and s together with 7, etc. are labelled cacuminal, and their phonetic differ-
ences are not revealed. Therefore, another coordinate of distinctive articulatory efforts
(prayatnam) is used to differentiate the large classes of sounds formed along the place of
articulation coordinate. Incidentally, it should be borne in mind that these coordinates and
their values are specific to Sanskrit speech sounds; they are not given by any general pho-
netic theory.

2. Articulatory efforts

There are two types of articulatory efforts: (1) Internal (abhyantar) and (2) External
(bahya). Internal efforts are made in the supraglottal region (from above the giottis to the
lips) and external efforts are made in the sub-glottal region of the vocal tract. Efforts refer
to modes of interrelation between articulators situated in the vocal tract.

Internal efforts are further subdivided into four: (1) contact (= sprstam) between the artic-
ulators; (2) slight contact (= isatsprstam) between the articulators; (3) openness (= vivrtam)
of the articulators; (4) closeness (= samvrtam) of the articulators.

External efforts are held to be eleven in number, (1) openness (= vivara) of the vocal cords;
(2) closeness (= samvara) of the vocal cords; (3) breath (= svasa); (4) vibration (= nada);
(5) voiceless (= aghosa); (6) voiced (= ghosa); (7) aspirated (= mahaprana), (8) unaspirated
(= alpaprana); (9) high pitch (= udatta); (10) low pitch (= anudatta), and (11) high-low
pitch (= svarita).

Incidentally, the first two external efforts (viz. openness and closeness of the vocal cords)
are physiological, the next two (viz. breath and vibration of the vocal cords) refer to the
acoustic output corresponding to the first two physiological efforts and the last two (viz.,
voiceless and voiced) refer to the linguistic correlates of the physiological and acoustic
efforts. That is, only the first two are articulatory efforts proper; the acoustic and lin-
guistic correlates are their results, not efforts in their own right. If this is accepted, the
number of external efforts would be reduced to seven from eleven.

395

However, whatever be the number of internal or external articulatory efforts, classification
of sounds by articulatory efforts is necessary for differentiating the large underdifferentiated
classes of Sanskrit speech sounds formed by place of articulation criterion. For example,
classification by place of articulation alone gives rise to the class of palatals including i; c,
ch, j, jh; y; s. This is an underdifferentiated class; it includes consonants, a vowel, as well as
a semi-vowel. But by using the internal articulatory effort of contact (= sprsta), we can
differentiate c, ch, j, jh from the remaining i; y; s. Next, the application of the internal
articulatory effort of slight contact (= isatsprsta) to this group distinguishes y from i and
$. y is said to be produced by slight contact between the articulators, while for i the artic-
ulators are held apart, as also for s. Thus, y and s are different from each other, because y
is characterized by the internal articulatory effort of slight contact between the articulators;
$, on the other hand, is produced by no contact, slight or full, between the articulators, a
condition which is necessary for the production of i as well. In other words, i and s are held
to be produced by one and the same type of internal articulatory effort, and since they are
produced at the same place of articulation in the supra-glottal region, they fulfil the condi-
tions for homogeneity as laid down by Panini's definition of homogeneity. Similar is the
case with a and h, r and s, / and s; the three pairs meet the conditions of homogeneity
stated by Panini's Savarna-rule. But from the view point of grammatical processes it is not
desirable that vowels and consonants be made homogeneous with one another. However, it
is unavoidable so long as the system of phonetic description and classification of Sanskrit
sounds by place of articulation and internal articulatory efforts remains unchanged. That
Panini himself was faced with the problem of preventing homogeneity between a and A, r
and s, / and s is evident from his rule ndjjhalau (1.1.10). But before concluding that Panini’s
solution is the only possible solution under the circumstances, it would be desirable to con-
sider at least one more solution, a phonetic solution of the problem.

3. Phonetic solution

It can be easily shown that the problematic homogeneity relation between a and A, r and s,
l and s is a consequence of limiting the number of internal articulatory efforts to only
four. The addition of one more type of internal articulatory effort, i.e., slight openness
(= isadvivrta) to the set of previously given four types would altogether rule out homoge-
neousness between the above mentioned. Then s, s, s would be produced by the articu-
latory effort of slight openness (= isadvivrta) between articulators as different from open-
ness (= vivrta) between articulators, which is the characteristic articulatory effort for a, i, r
and /. Indeed, in his extensive commentary on 1.1.9 in Mahabhasya, Patafijali has discussed
this possibility of increasing the number of internal articulatory efforts to five. And it
should be noted that it is not an ad hoc manoeuvre proposed to solve a nagging problem
regardless of its bearing on the system of phonetic description and classification. In fact the
proposed internal articulatory effort of isadvivrta is logically implied by the widely ac-
cepted, canonical articulatory effort of isatsprsta: slight contact implies slight openness.
Thus, it is obvious that the inclusion of isadvivrta in the class of internal articulatory ef-
forts precludes the possibility of undesirable homogeneity relation between consonants and
vowels. And it is equally obvious that Panini did not include isadvivrta in the class of
internal articulatory efforts. Had he done it, he would not have had to formulate the rule

396

najjhalau (1.1.10). Panini accepted the traditional system of phonetic description and
classification in its entirety and faced the problem inherent in it.

4. Paninean solution

Panini solves the problem of undesirable homogeneity between a and A, i and $, r and
s, J and s by formulating an exception to the savarna-rule 1.1.9. The exception rule is
najjhalau (1.1.10). It prohibits vowels and consonants from being homogeneous with one
another. Accordingly, homogeneity established on articulatory phonetic grounds between
à, i, r and / and A, S, s, and s respectively by 1.1.9 is prohibited by 1.1.10.

The simplicity and finality of the solution is truly Paninean and is in accordance with the
organizing principle of his grammar. The principle is that if there is incongruity between
phonetic and phonological frames of description; if articulatory and linguistic classifica-
tions do not coincide, let the phonological and linguistic prevail over the phonetic. There-
fore, let the sounds which turn out to be homogeneous in phonetic terms be not treated as
homogeneous for grammatical purposes. This strategy allows Panini to accept the articulatory
phonetic classification of sounds transmitted to him by tradition, but without its undesir-
able consequences. Tradition is preserved, even developed, since the limitations inherent in
its system of phonetic description and classification are seen for what they are and are
overcome.

However, what is of greater significance here, and needs stressing, is the technique and the
terms by means of which Panini overcomes these limitations. najjhalau (1.1.10) is an apt
illustration of his technique and terms. Panini uses the technique of pratyahara for forming
classes of Sanskrit speech sounds which are appropriate operands of high- order grammat-
ical rules. That is, the principle that governs and regulates the classification of sounds in
Panini’s grammar is not articulatory phonetic but linguistic and grammatical.

The rule najjhalau (1.1.10) contains two pratyaharas: ac and hal. Actually ac constitutes
the class of all Sanskrit vowel sounds, viz., a, i, u, r, L, e, o, ai, au; hal constitutes the class
of all Sanskrit consonants and semi-vowels. That is, ac and hal are two mutually exclu-
sive classes of sounds which have different contextual functions in the language. The
technique of pratyahara enables one to form as many sub-classes of these general classes as
are necessary for describing the language with economy and generality. For example, the
ac-class can be further sub-divided into smaller classes: ec (7 e, o, ai, au); aic (7 ai, au);
ak (2a, i, u, r, D;ik (7i u, r, Duk (9u, r, D), etc. Similarly, the Aal-class can be sub-divided
into several sub-classes of consonants like: jhal (= jh, bh, gh, dh, dh, j, b, g, d, d, kh, ph, ch,
th, th, c, t, t, k, p,$, 5,5, h);jas (= j, b, g, d, d); khar (= kh, ph, ch, th, th, c, t, t, k, p, $, s, s);
etc. The point is that the technique of pratyahara is extremely useful in forming classes and
sub-classes of sounds which are needed as operands for grammatical rules. Another feature
of these classes so formed is that they are not articulatory phonetic classes; they are ex-
tensional classes of sound segments whose only justification seems to be the fact that they
help in an economical and compact description of the language with maximum generality.

This is obviously different from the technique that Panini uses in defining the class of
homogeneous sounds in articulatory phonetic terms. And, as has been described earlier, the
definition constituting the class of homogeneous sounds is not free from serious defects;

397

for instance, it leads to the grouping of consonants and vowels as homogeneous. The defect
is eliminated by the grouping of sounds on the basis of pratyahara technique. That is, the
intensional definition of the class of homogeneous sounds sanctioned by the traditional
articulatory phonetic criteria gives rise to a problem; the problem is solved, however, in
terms of the extensional classes of sounds formed by the powerful technique of pratyähara.
Evidently, Panini does not develop the technique only to prevent the undesirable homo-
geneity between a and À, i and Ss, r and s, / and s; indeed, his grammar of Sanskrit makes
only a marginal use of homogeneity principle and homogeneous classes, as constituted by
1.1.9 but the sound classes produced by pratyaharas form the basis of his system of descrip-
tion in Astadhyayi.

The preceding discussion has brought to light the difficulties that the definition of homo-
geneity in terms of internal articulatory efforts gives rise to, and the solution that Panini's
grammar offers to these difficulties. However, the application of some of the savarna-rules
for the derivation of correct grammatical forms brings to light yet another inadequacy of
the definition of homogeneity by 1.1.9.

Consider the derivation of utthanam (‘rising up’). The initial appropriate rule for the pur-
pose is udah sthastambhoh pürvasya (8.4.61). Briefly, the rule states that given the se-
quence ud + stha or ud + stambha let s of stha and stambha be replaced by the sound
which is homogeneous with d of the preceding segment ud. However, the selection of the
appropriate sound homogeneous with d must be conditioned by s of stha or stambha. In
effect then, s is to be replaced by a sound which is homogeneous with both d and s. Now,
in respect of the place of articulation 7, tA, d, dh and n are all articulated at the same place
in the supraglottal region where s is articulated. But in respect of internal articulatory
efforts, they are all different from s; while s is articulated with openness between artic-
ulators, they are all articulated with contact between articulators. Therefore, none of them
satisfy the condition of being homogeneous with s.

Nevertheless, th is deemed to be homogeneous with s and is selected to replace s in ud +
stha or ud + stambha. This is done apparently by making external articulatory efforts, too,
criterial for homogeneity among sounds. Only by this manoeuvre are th and s held to be
homogeneous: s by its place of articulation is dental, so is th; s by its internal articulatory
effort is open, but th is not; however, by its external articulatory effort s is voiceless and
aspirated, so is th. Therefore, s and th are considered homogeneous with each other, in
spite of the fact that their internal articulatory efforts are quite different. Thus, the se-
quence ud + stha by udah stha- stambhoh purvasya (8.4.61) is first replaced by ut + ththa
by khari ca (8.4.55), and the output of 8.4.55 is replaced by ut + tha by jharojhari savarne
(8.4.65), and it in turn is replaced by uttha + lyut by lyut ca (3.3.115), and next by
utthana by yuvornakau (7.1.1.) and finally by utthanam by atoms by 7.1.24. Similarly,
we can derive uttambhanam. This derivation shows that external articulatory efforts are as
essential to homogeneity among sounds as are internal articulatory efforts, although 1.1.9.
does not consider them for the definition of homogeneity.

One more example, which leads to the same conclusion, is given below. Consider the deri-
vation of vaggharih or vagharih according to the rule: jhayo ho anyatarsyam (8.4.62). The
rule states that given the consonants of jhay -- pratyahar (excluding the voiceless aspirates
and unaspirates) immediately preceding A, h may be optionally replaced by that which is
homogeneous with the sound preceding h; however, the replacing sound must be condi-

398

tioned by h and be homogeneous with it also. For example, let the string be vac + hari >
vak + hari by choh kuh (8.2.30) ^ vag * hari by jhalam jasonte (8.2.39). Now, in this string
g immediately precedes h; the sounds homogeneous with it are k, kh, g, gh, n, but none of
them is homogeneous with A in respect of the internal articulatory efforts, because they are
all produced with contact between the articulators, while À is produced with the openness
of articulators. Therefore, the selection of the proper replacement of h from among k, kh,
g, gh, hi is blocked. If, however, external articulatory efforts are also considered essential to
the selection of the right replacement for h in the string, gh fulfils the conditions of being
homogeneous with h. Naturally, gh is selected as the replacement of h, since gh is voiced,
aspirated and is produced with the closeness and vibration of the vocal cords, just as A is
voiced, aspirated, and is produced with the closeness and vibration of the vocal cords. Only
then can vag + hari be replaced by vagghari by jhayo ho anytarsyam (8.4.62) or optionally
remain as vaghari. By certain other sup-affixes rules we can derive vaggharih or vagharih.
The point here is simple: 1.1.9 defines the class of homogeneous sounds in terms of internal
articulatory efforts, but the application of rules stated in terms of homogeneity leads to
certain difficulties, which are removed only if external articulatory efforts are considered
necessary for the purpose of defining homogeneity between sounds. That is, the definition
of homogeneity by 1.1.9. is at variance with the application of homogeneity in the deriva-
tion of correct grammatica! forms.

Earlier on we noted how by itself tulyasyaprayatnam savarnam (1.1.9) led to undesirable
homogeneity between consonants and vowels, but how by najjhalau (1.1.10) they were
declared non-homogeneous, and thus was removed the difficulty that 1.1.9 had given rise
to. Now we notice that in the derivation of certain grammatical forms like vaggharih and
utthanam the principle of homogeneity alone is not sufficient; the homogeneity principle
stated as it is in terms of internal articulatory efforts is not adequate. However, there does
not seem to be any rule in Astadhyayi explicitly stated to remove this specific inade-
quacy — one is left to an ad-hoc strategy of using external articulatory efforts also as
essential to the establishment of homogeneity between sounds. Or else to use sthane antar-
tamah (1.1.50) to support the inadequate definition of savarna for the purpose of deriving
utthanam, etc. In short, neither the definition of savarna nor the operation of savarna-
rules is free from deficiencies. The reason may very well be that Panini has defined savarna
in phonetic terms.

Two Anatolian forms: Palaic afkfumauwa-, Cuneiform Luvian wa-a-ar-Sa

Calvert Watkins (Harvard University)

1. Palaic aXkummauwa- ‘holy, consecrated, pure, tabu’

This form is attested four times in two texts: KUB XXXV 165 Rs. 11 and KBo XIX 152
I 14 with its duplicate KBo XIX 153 III 9, 20 (restored). They are easily accessible in
edited form (Carruba 1970; cf. also Carruba 1972, A. Kammenhuber, RHA 64.1353; BSL
54.1353 :18—45).

The first of these tablets (Carruba 2A) is written in the Old Hittite script; it gives lengthy
Palaic cultic utterances in an Old Hittite ritual, i.e., a bilingual context. The relevant
passage reads:

nu UZ Ušu-up-pa ti-an-zi nu me-ma-i nu-ku Dza-par-wa az aS-ku-ma-a-u-wa-ga
wa-aq-qa-kan-ta

[Hittite] ‘They place the consecrated meats and he speaks as follows: [Palaic] “And
now, Zaparwa, these consecrated meats are waq?kant-.”””

The Palaic prayer to the god Zaparwa continues with an enumeration of the meats which
are aXkumduwa- and then the instruction “Give them to the goddess Katahzipuri!” ne
Dka-tah-zi-wu,ri piisa]. In this passage afkumauwalga) is clearly the Palaic equivalent of
the Hittite cultic term UZ Ux, ppa. Like the latter (and its Umbrian cognate sopa supa) it is
neuter plural, as proved by the concord with the participial wag?°k-ant-a and the anaphoric

pronominal n-e.!

The second example of our word is found in a prayer (ma/te$Sar) in another ritual (Carruba
2A), where it begins a paragraph immediately following the sacrificial ‘carmen’ which I
have treated (1979:305—314; 1972). The sentence reads, as Carruba rightly translated,
aX-ku-um-ma-u-wa-a3 ha-a-an-ta ti-i-li-la ha-a-ri ‘The consecrated foods (n.pl.com.) are hot
(3 pl.mid.), the gi (n.pl.nt.) are hot (3 sg.mid.).’ The duplicate offers the spelling variant
[a$-K]u-ma-a-u-wa-a$. The gender and number agreement follow the archaic Indo-European
syntactic rule. On the (inherited) verb cf. Carruba (1970:53), Oettinger (1977) and Mel-
chert (KZ 97. 1984:24).

Kammenhuber (1959a: 56) would render the passage as dem kultisch reinen (Fleisch) ge-
gegenüber plaziert er die filila^, taking aYkummduwas as dative plural and hanta as postposi-
tion; she is followed by Szemerényi (1979:315ff). But the scriptio plena in ha-a-an-ta
(which Szemerényi ignores) forbids identifying it with Hitt. (menah)handa or OHitt.
(man)handa. The duplicate 3B offers ha-an-da twice, to be sure; but this manuscript
regularly shows younger spelling habits than 3A, like omission of plene-vowels and sub-

490

stitution of logograms for syllabic spelling: the authority of 3A’s ha-a-an-ta must be recog-
nized.” That a semantically differentiated animate plural aSkummauwa§ designating some-
thing ritually pure to cat could coexist with the inanimate a¥kumawa(-ga) = UZU3uppa
‘ritually pure meat’ is, pace Szemerényi, entirely possible; cf. MINDA šuppi ‘ritually pure
bread or porridge’ KBo X 3415, and Hoffner (1974: 184—5 and 211).

The first breakthrough in the interpretation of this curious-looking word was made by
Szemerényi in a footnote (1979:314 n. 6). Rather than seeing in aSk- a spelling of a cluster
[sk], with Carruba (1970:39), Szemerényi (1979:316) takes the Palaic word as ‘a com-
pound (as + kuma-) expanded with the suffix -wa-’. The second element or ‘central part’
-kuma- (rectius -kumma-) he brilliantly equates with Cuneiform Luvian kumma- (rectius
kummai-) ‘sacré, pur, tabou’, a word identified by E. Laroche and H.G. Giiterbock (Ugari-
tica 3:154; cf. OLZ. 1959:275) and Laroche 1959 6.56—57; Laroche (CRAJ 1974: 124)
further saw its connection with the family of Lycian kumaza ‘priest, hiereus’ and related
words on the trilingual Xanthos stele discovered in 1973. Note also a Hieroglyphic Luvian
ku-ma-za-sa Kayseri 5C sentence 16. A full discussion of the Luvian and Lycian evidence
for the forms can be found in Laroche (Fouilles de Xanthos 6.1979: 68—110).

The stem kumma- is the semantic equivalent of Hittite Suppi- and UZUfuppa ‘sacré, pur,
tabou; holy, consecrated (to the gods), tabu (for human beings)’; it is thus precisely the
opposite of Szemerényi’s ‘kosher’. For the notion in Anatolian, compare Hitt. Siuni-mi-ma-
mu kuit Xuppi adanna natta ara n-at UL kuššanka edun ‘But what is sacred to my god (and)
not right for me to eat, I never ate’ (KUB XXX 10 obv. 10, the prayer of Kantuzilis). Now
once we accept Szemerényi's analysis of our word as a compound with second member
kumma-, the problem is reduced to identifying the first member. I propose to see in Palaic
as- the correspondent of Cun. Luv. a-a-aS- and Hittite a-i-i§ (ai§) ‘mouth’, Hier. Luv. a-sa5-za-
“speak”?

The literal sense of the compound would then be something like “mouth-tabu.” Such a
compound has a non-Indo-European ring to it (though recall English foofsore, heartsick,
headstrong, Armstrong); syntactically it is very similar to Hittite pattar-palhi- ‘wing-broad’,
and like the latter could have been calqued on a non-Indo-European language of the region.
But the components of the compound are inherited. For the form of the compound, we
may either follow Szemerenyi’s analysis (as + kumma) + wa-, with the last a secondary
suffix, or we may segment as + kummawa- and assume kummawa- is the Palaic word for
‘sacred’, with a different suffix from that of Luvian kummai-.

Most of the etymological literature on the Anatolian family of Palaic a$- ‘mouth’ (to Lat.
Os etc.) is conveniently summarized in Tischler (1982 s.v. aiš/iš-). Lindeman (1967:1188—
90) reconstructs h3ehj-es, while Szemerényi (1979b: 267-83) prefers *o-es. In my view a
prototype *(h1)eh3-es (like *nebh-es > Hitt. nepiš), whence automatically *(hj)oh3es
by laryngeal coloring within morpheme boundary, yielded Common Anatolian *aé3. This
gave Hittite aif (OH a-i-i$), either with hiatus ais, or ais by a glide formation rule; cf. OH
(-a)-a-i-iz-zi < *-deti < *-gjeti.*

The Common Luvian *e > a rule produced “ads, which was probably contracted to as,
though the spelling a-a-a¥ could also render the hiatus form. The existence of an *e >a rule
in Palaic however, as commonly assumed, is far less certain that in Luvian. But in view of
the imperative 2 and 3 singular of the verb ‘to be’ a-a5, a-a¥-du (Hitt. e-e$, e-eX-du) we are
justified in assuming a stem “ads in Pre-Palaic, whence attested aš.

401

The origin of the word kumma- is disputed. Laroche (Actes du colloque sur la Lycie anti-
que 2.1980) takes both kumma- and suppi- as ‘substratal’, but the latter is surely inherited,
cf. note 1 above. Szemerényi (/nL 4.1978: 176) has proposed connection of kumma- to the
Avestan root noun sü- ‘prosperite’ (Kellens 1974) and the family of Vedic süra- savistha-,
Av. süra- savifta- ‘powerful’, Greek kúrios ‘authoritative’, ákūros ‘without authority”. We
would have a Caland-like *kuhx-mo-, plus further suffixes; if Doric pepamai ‘possess’
belongs here the laryngeal is #2. This etymology would be valid only on condition that
*hom- gave Common Anatolian -mm-, which is uncertain. J. Schindler suggests rather (oral
communication) a derivation kumma- < *kun-mo- with zero grade of the *kuen- of Avestan
span-ta- ‘holy’, Lith. Svefitas, OCS svéte, which is semantically and formally more attractive.
We are almost certainly dealing with a Common Anatolian vocable; Hittite kunna- ‘right;
favorable’ doubtless contains the same root, as H.C. Melchert points out to me.

That kumma- is in fact Common Anatolian is indicated above all by its geographical range
as an element in place-names, which seem not to have been considered in this connection
in the literature.’

The element kumma- is surely to be identified in the local names URUKumma, URUKum-
maha (= classical Kommagéné), in (KUR)URUKummanni, the great religious center in
Kizzuwatna in the southeast (classical Comana Cappodociae), and also in URUKummiya
‘the Storm God’s residence, somewhere in the mountains bordering Mesopotamia to the
north’ (Güterbock 1951: 78). All of these must mean something like ‘Holy Land’, ‘Holy
City’; and the Greek name of the Phrygian city of Hierapolis in western Anatolia suggests
a continuous and conscious semantic tradition going back at least a millenium. The resi-
dences of the gods were ‘holy’, and a new fragment of the Song of Ullikummi provides a
striking collocation: KBo XXXVI 65 IV 26—8 nepiSi-wa-kan Sa LUGAL-iznanni paimi
nu-wa-za URUKummiyal- Jaa Suppa EMESDINGIRMES Efkluntarrann-a dahhi
DINGIRMES ma-wa-kan ne[(piSaz ka})\tta memal GIM-an ishuwahhi ‘I will go up to the
kingdom of heaven, and I will take for myself the city of Kummiya and [ |, the holy
temples and the kuntarra- house, but the gods I will shake down from heaven like meal."
I think it quite possible that kumm-iva/kummana were etymologically transparent to the
Hittites.

Note finally, from yet another part of Anatolia, the river IDKummitmahaX and URUKum-
maneSmah. Since at one time this river formed the border between Hittite and Kaska
territory, it must have been north of Hattusas, probably northeast (Schuler 1965: 19, n.2).
The name Kummišmaha is evidently a compound with first member ‘Holy’; the second
member is conceivably related either to the name of the cult city Samuha, apparently to
the northeast in the Upper Country, or to the name of the city Sammaha, to the northwest
in the direction of Pala, or to both (Garstang-Gurney 1959:31; Schuler 1965:46).

2. Cuneiform Luvian wa-2-ar-Sa

Since Meriggi (WZKM 53. 1956, 204-217) and Laroche (1959:152) this word has been
identified with Hittite warSa- in the meaning ‘drop’; Meriggi (OLZ 1962.259 n.1) further
suggested the derived meaning ‘teardrops, tears’ for one obscure passage. The equation is
however less than perfect, since the Luvian word apparently is neuter plural and the Hittite
animate, and since the semantic orientation of the Hittite word toward water as a weather

402

manifestation, ‘cloudburst’ (IE *uers-), is absent in Luvian. The one clear context, more-
over, in a Hittite-Luvian ritual favors the meaning ‘water’ pure and simple, and the mor-
phological concord there points unambiguously to a neuter singular.

Laroche in his edition of the passage in question (1959: 152) significantly entitles it “Con-
juration de l'eau et du sel". KUB XXXV 54 III 12—16 gives the 'directions' in Hittite:

nu-%an ANA GAL GIRq kuit watar lahuwan MUN-ya-kan anda iShuwan n-at-kan E-ri anda
pappar$zi ANA BEL SISKUR.SISKUR-ya-San papparszi nu kik$[an m|emai

‘The water which is poured into the asphalt bowl and the salt which is shaken in it, he
sprinkles in the house, and sprinkles on the master of the ritual, and he speaks as follows:’
[17-21 Luvian]

[w]a-a-ar-Xa-at-ta ÍD-ti [na-n ]a-am-ma-an

[MJUN 3a-pa a-a-la-a-ti u-wa-a|(-ni-ya)-ti] 4-pa-am-ma-an
[w]a-a-ar-Xa-at-ta zi-i-I[a ÍD-i] an-da

[n]a-a-wa i-ti MUN-Sa-pa-a[t-ta z}i-la

[(a-a-)Mi-i u-wa-a-ni-ya na-a-[wa i-t]i

‘The water is led from the river,

and the salt is brought from the steep rock face;

the water will go henceforth into the river no more,

and the salt will go henceforth to the steep rock face no more."?

It seem obvious to me that wa-a-ar-Sa (-tta is an enclitic particle) corresponds to the Hittite
watar ‘water’, just as MUN-Sa (-pa is an enclitic conjunction) corresponds to the Hittite
MUN ‘salt’. Furthermore the broken passive participle -amman is clearly neuter singular;
wa-a-ar-Sa cannot be neuter plural. The solution is clear: The word for ‘water’ is a neuter
noun wa-a-ar, suffixed by the so-called ‘-3a case’, like MUN-Sa and all the following neuters
in the passage:

(22-3) adduwal-za utar-Sa halliS-Sa parattan-za

‘evil word, impurity, uncleanliness.'

This possibility was indeed seen by Carruba in his study of the Luvian -3a case (n. 3 and 7
supra), but he maintained the meaning ‘Tropfen’, for which not surprisingly the stem was
to him "nicht klar." But once we recognize the meaning as ‘water’ we can exactly equate
Luvian wa-a-ar with Vedic var nt. ‘water’ and the animatized Avestan var- m. ‘rain’. See in
general Pokorny (1959:80), and Mayrhofer (1976: 194). Ved. var is attested 10 times in
the Rigveda, three in the family books. Two of these three show disyllabic vaar, as does
also one clearly archaizing or archaic hymn of book X, and perhaps one instance in the
Atharvaveda (IV 7.1, cf. Whitney ad loc.). The disyllabic variant is therefore certainly older
and linguistically real; we must reconstruct an old r-stem *uehj-r. Luvian wa-a-ar- reflects
just this preform; as in a-a-a§- ‘mouth’ the scriptio plena renders either a long vowel (by
contraction) or the hiatus form waar. The expected zero-grade *uh -r- appears in ON ur
‘drizzle’ and in Lat. urino(r) ‘dive, mergi in aquam (Varro)’ from urina “*water’; the latter
is animatized by the same feminine suffix as reg-ina.

In the Rigveda we find var used figuratively of ‘milk’; Oldenberg (1912) ad IV 5.8 yád
usriyanam dpa var iva vrán ‘which they have uncovered like the “water” (= milk) of the
dawn-cows’. Geldner ad loc. and Renou (EVP 13/3.1964:37) take var as ‘cover’ or ‘secret’;
but the secret at issue is the hidden caru prsneh ‘the precious {udder (Geldner), or milk

403

(Renou)] of the spotted cow Práni'. Compare X 12.3 duhé yád éni divyám ghrtám vah ‘that
the speckled hind give heavenly ghee (and) milk’. The cadence ghytam vah recurs at X 99.4.

This metaphor may be older still. For as I hope to show at greater length elsewhere, to
Vedic var ‘water; milk’ corresponds Archaic Old Irish fir ‘milk’. The Irish word is a dif-
ferent one from fir ‘true’; it is glossed as mblicht (bd) ‘(cow’s) milk’ and finn ‘white’ by the
native tradition. The word occurs in a single formulaic phrase teora ferbba fira ‘three milch
cows’ (accusative plural), preserved independently in two mythological-legal traditions
about dawn-cows: the beginning of the law text Cetharslicht Athgabala (AL 1.64; CIH
325.25, 881.4, 1897.16) and a judgment of Fachtna mac Sencha in Cormac’s Glossary
§ 585. Archaic Irish fir reflects a substantivized Indo-European possessive adjective *ueh
-r-O- ‘Shaving water, aqueous’, from *uehj-r; it allows us to specify the laryngeal of this
root. Semantically the passage is ‘watery’ > ‘milky’ > ‘milk’. Here fira is probably genitive
singular (hypostasized to u-stem inflexion, like fir ‘truth’), and Kim McCone plausibly
regards ferb fíra as semantically equivalent to Mod. Ir. b6 bhainne ‘milch-cow’ (personal
communication). The extraordinary archaism which fir is owes its preservation to being
frozen in a formulaic epithet in a mythological nexus which is itself of Indo-European
antiquity.

Notes
1 On the segmentation and interpretation of afkumäuwa-ga ‘these aSkumduwa’ | follow the excellent
suggestion of H.C. Melchert (XZ 97.1984:25). I therefore withdraw my earlier explanation (1982:
256) of this -ga as a reflex of word final ^5 of the neuter plural (Watkins 1975/6), likewise that of
the -ha of Pal. wa-su-u-ha, correctly seen by Meichert to be wasu + ha, ibid. 31—3. On the other hand
I do continue to maintain that the -ga- of verb forms such as tak-ku-wa-a-ti/tak-ku-wa-ga-ti is a
reflex of h).

Compare in the preceding five lines alone:

A Sa-me-i-ri-if B Xa-me-ri-i¥
ha-la-a-i8-ta ha-la-i$-ta
wa-a-aX-hu-ul-la-ti-ya-a* wa-aX-hu-ul-la-ti-ya-aY
ha-a-ap-na-aS-ta ID-an-af-ta
ha-Si-i-ra-am-pi GIR-an-pat
pa-da-a-am-ma-an pa-ta-am-ma-an
3 For the Cuneiform Luvian forms see Laroche (RHA 76.1965:45); for the Hieroglyphic Luvian see
Hawkins and Davies (RAS 1975:131-133). CLuv. a-a-at-Xa(-) shows the - ‘case’, a possibility seen
by O. Carruba in his valuable study of the latter (1982:1—15, esp. 5).
^ [s the i- of the Hittite oblique stem iS$- a prothetic vowel, before the original cluster of zero-grade
root and suffix, in an amphikinetic neuter paradigm nom.-acc. *(/1)eh3-(e)s, gen. *(h1)/3-s-es?
5 Laroche (Fouilles de Xanthos 6.1979:109 n. 42) does rightly note that Kummani, Comana must
mean ‘ia Sainte’, comparing Semitic Qadesh et al. References for the local names may be found in
à H. Ertem (1973) and G.F. del Monte and J. Tischler (1978) s. vv.

Cf. Güterbock (1952) 48. The new fragment confirms most of the EE s restorations. Col. III 22 of
the same tablet (XXXIII 106) offers 1 x-aY Yuppa f [MES DINGIRMES in broken context.

O. Carruba in his version of this passage, Serta Indogerm., Festschrift G. Neumann 47—6 (1982),
translates the first phrase in the traditional way following Laroche (Les gouttes sont [...] ées du
fleuve) as ‘die Tropfen sind vom Flusse ge-.. .-t’, though in Gedenkschrift Kronasser 5 (1982) he
envisages another possibility. Hittite war$a$ in any case must mean ‘cloudburst, downpour’ (so
already Friedrich, HWb.) not ‘drop’, as Giiterbock points out to me.

For the restoration [nan]amman ‘led’, which fits the space and the traces, compare nandtti (dupl.
lalátti ‘takes’), nananta DLL. Luvian wear nana- ‘aquam ducere’ paralles Hittite watar nai-.

404

For ali- ‘steep, high, sheer’ and /u)wani- ‘rock face, block of stone, stele’ (Hier. wa-ni-za) see “The
language of the Trojans” in the Proceedings of the Bryn Mawr Symposium on the Trojan War, ed.
Machteld J. Mellink.

For the provenience of salt from mountain rock faces in Anatolia cf. Güterbock’s comments cited by
me in “Questions linguistiques palaites et louvites cuneiformes”, in the proceedings of the Colloque
Anatolien, Paris 1985 (ed. R. Lebrun: to appear).

References

Arnold, Edward Vernon, 1905. Vedic metre in its historical development (Cambridge: University Press).

Carruba, Onofrio, 1970. “Das Palaische” (= StBoT 10) (Wiesbaden: Harrassowitz).

Carruba, Onofrio, 1972. Beiträge zum Palaischen (Wiesbaden: Harrassowitz).

Carruba, Onofrio, 1982a. “Beiträge zum Luwischen”, in: Serta Indogermanica. Festschrift für G. Neu-
mann (Innsbruck: Institut für Sprachwissenschaft der Universität Innsbruck) [Innsbrucker Beiträge
zur Sprachwissenschaft].

Carruba, Onofrio, 1982b. “Der Kasus auf -Ja des Luwischen”, in: Investigationes Philologicae et Com-
parativae (Gedenkschrift Kronasser), edited by E. Neu (Wiesbaden: Harrassowitz), 1--15.

Ertem, Hayri, 1973. Cografya Adiarí Dizini (Ankara: Universite).

Garstang, John/Gurney, O.R., 1959. The Geography of the Hittite Empire (London/ Ankara: Brit. Inst.
Inst. of Archaeology).

Güterbock, Hans Gustav, 1952. The Song of Ullikummi (= JCS 5-6).

Hoffner, Harry A., 1974. Alimenta Hethaeorum. Food production in Hittite Asia Minor (New Haven,
Conn.: Am. Oriental Soc.).

Kellens, Jean, 1974. Les noms-racines de l’Avesta (= Beitr. zur Iranistik 7) (Wiesbaden: Reichert).

Laroche, Emile, 1959. Dictionnaire de la langue louvite (Paris: Adrien-Maisonneuve).

Lindeman, Frederik Otto, 1967. “Indo-européen *6s ‘bouche’”, in: To honor Roman Jakobson 2,
(= Janua Linguarum Ser. Maior 32) (The Hague/Paris: Mouton), 1188—1190.

Mayrhofer, Manfred, 1976. Kurzgefafites etymologisches Wörterbuch des Altindischen 3 (Heidelberg:
Winter).

del Monte, G.F./Tischler, J., 1978. Die Orts- und Gewässernamen der hethitischen Texte (Beihefte zum
Tübinger Atlas des Vorderen Orients B 7/6) (Wiesbaden: Reichert).

Oettinger, Norbert, 1977. Die Stammbildung des hethitischen Verbums (= Erlanger Beiträge zur Sprach-
und Kulturwissenschaft 64) (Erlangen: Nurnberg).

Oldenberg, Hermann, 1912. Rgveda. Textkritische und exegetische Noten (Berlin: Weidmann).

Schuler, Einar von, 1965. Die Kaškäer. Ein Beitrag zur Ethnographie des alten Kleinasien (Erganzungs-
band zur ZA, N. F. 3) (Berlin: de Gruyter).

Szemerenyi, Oswald, 1979a. “Palaic and the Indo-Furopean Laryngeals”, in: Florilegium Anatolicum
(Mélanges offerts a Emanuel Laroche) (Paris: E. de Boccard), 315-319).

Szemerényi, Oswald, 1979b. “On Reconstruction in Morphology", in: Studies in honor of Archibald
A. Hill 3, edited by Mohammad Ali Jazayery et al. (The Hague/Paris/New York: Mouton), 267 —283.

Tischler, Johann, 1977. Hethitisches Etymologisches Glossar (Innsbruck: Institut für vergleichende
Sprachwissenschaft).

Watkins, Calvert, 1972. Indo-European Studies (Harvard: Dept. of Linguistics).

Watkins, Calvert, 1975a. “Die Vertretung der Laryngale in gewissen morphologischen Kategorien in den
indogermanischen Sprachen Anatoliens” in: Flexion und Wortbildung, edited by H. Rix (Wiesbaden:
Reichert), 358—378.

Watkins, Calvert, 1975b. **La désignation indo-européenne du 'tabou'", in: Langues, Discours, Société.
Pour E. Benveniste, edited by J.C. Milner (Paris: Editions du Seuil), 208—214.

Watkins, Calvert, 1979. “A Palaic Carmen”, in: Studies in honor of Archibald A. Hill 3, edited by M.A.
Jazayery et al. (The Hague/Paris/New York: Mouton), 305—314.

Old Indic sünu-, Greek huius ‘son’

Werner Winter (Universität Kiel)

Szemerenyi (1977: 10-12) provides a convenient overview of currently held opinions on
the oldest Indo-European terms for ‘son’. He sides with earlier views that Olnd. sunu-, Lith.
sunus, OCS syniit clearly presuppose an antecedent form *sunus. Avest. hunu- he takes to
represent Olran. Aunu-; the Germanic forms with short vowel of the first syllable (Goth.
sunus, etc.) he believes, along with Dybo, to have resulted from ‘secondary shortening in
unaccented position’. PIE *sunu-s he considers (1977: 11) ‘a clear derivative of the verbal
root su- (: Ind. sw-te “gebiert”), which originally meant “offspring” but was later restricted
to the sole important offspring, the male’.

He thinks that *sunu-s was ‘an all-IE word’, and that the competing form found, e.g., in
Gk. huitis ‘son’, was the result of an innovation (Szemerényi 1977: 11, 52). Without really
committing himself, he appears to want to derive the Greek form from *su-yu-s and tends
to think that this reconstruction finds support in Tocharian B soy, A se, adding, however,
the cautionary comment (1977: 11) ‘if these do in fact derive from *suyu-, and are not
borrowed’.

Elsewhere (Winter 1985: 261—263) I have presented arguments why the Tocharian and the
Greek forms should be derived from PIE *swXyu- (with different vocalizations of the first
syllable in Greek and Tocharian, respectively); as to Greek, I pointed out that it seemed
impossible to assume retention of plain intervocalic *-y- of a supposed pre-form *su-yu-.

In view of the fact that earlier scholars had connected u- in Arm. ustr 'son' directly with
Gk. Auius (cf., e.g., Walde — Pokorny 1927: 470 with references), it is somewhat surprising
that Szemerényi fails to mention the Armenian term. Obviously, an agreement between
Greek and Armenian is not unexpected (cf. Solta 1960: 462—466 with references); the
ties between Greek and Armenian, on the one hand, and Tocharian, on the other, are,
however, even weaker than Solta (1960: 477) is inclined to believe — none of his proposed
isoglosses will bear closer scrutiny. It is interesting to note that Van Windekens, while
listing (1976: 615—616) more than seventy supposed agreements between Tocharian and
Greek alone (the vast majority of which though cannot be upheld once one applies strict
standards), contents himself (1976: 619) with enumerating just four items which are
supposed to show an agreement between Armenian and Tocharian (none of which is found
in Solta 1960: 460-461, and none of which is in any way persuasive). The findings of
Adams (1984: 396), in his own evaluation, do not seem to be compelling enough to claim
a particular closeness between Armenian and Tocharian. There thus appears to be insuf-
ficient evidence to support the assumption that Greek, Armenian, and Tocharian owed
their matching forms for 'son' to shared innovation.

406

Traditionally, two different approaches have been chosen to explain *-nu- and *-yu-: either
they were taken to be indivisible formatives (as, e.g., by Brugmann 1906: 290, 224) or se-
quences *-nu- and *-y-u- (cf. Walde — Pokorny 1927: 469). Both analyses are beset with
difficulties:

A morpheme *-yu- is not attested outside the very word for ‘son’ in Greek, Armenian, and
Tocharian (except that Brugmann 1906: 224 is able to adduce the adjective Gk. praus
‘kind’); where *-yu- can be isolated in forms based on verbs (agent nouns, to be sure), as in
Indo-Iranian and Balto-Slavic, ‘son’ is formed with *-nu- and not with *-yu-. There is, thus,
no sound reason to consider *suyu- the result of an innovative reshaping of *sunu-: if there
was no *-yu- that could have been transferred, no remodeling could happen, be it in the
presumed group in its entirety or in each language by itself.

The bimorphemic interpretation of *-n-u- and *-y-u-, cited above from Walde — Pokorny
(1927: 469), has an antecedent in Brugmann (1906: 223) as far as *-y-u- is concerned:
*-y- is equated with the formative *-ye-/-yo- found in OInd. suyate; obviously, it is then
only natural to identify *-n- in *-n-u- with the nasal affix of Avest. hunami.

*.u- suffixes are not uncommon in deverbative adjectives (cf. Brugnann 1906: 176-179),
and there are indications that not only plain, but also paradigmatically modified, stems
could be used in such adjectives; thus, not even the tie-up with Avest. hunami needs to be
rejected outright. Also, the difference between adjective and noun is not necessarily crucial.

Much more troublesome, though, is another point: whenever derived from a transitive verb,
*-u- adjectives seem to have an active meaning; thus, if PIE *su- meant ‘beget, give birth’,
*sunu- and *suyu- could not have been used to denote ‘birth’ or ‘offspring’, but only
something with an active meaning. If this is the case, a masculine form *sunus or *suyus
would be an excellent descriptive term for ‘father’, but most certainly not for ‘son’.

The discussion so far can be summed up as follows: There is no comparative evidence that
would allow us to relate *sunu- and *suyu- as derivatives with a verb base meaning ‘beget;
give birth’. Furthermore, there is no support for the assumption of an innovative status of
*suyu-: therefore, *sunu- and *suyu- should be viewed as synchronically competing terms
in reconstructed Proto-Indo-European.

A few simple observations seem in order here: Whatever our interpretation of *-nu- vs.
*-yu-, we have to recognize that the two must be assigned a function or meaning so close to
being identical in both instances that they can be called (near-)synonymous. Furthermore,
it is reasonable to assume that *sz- as found in *sunu- is not different in meaning from *su-
in *suyu-. *sunu- thus can be said to be analyzable as ‘A’ plus ‘X’, and *suyu- as ‘A’ plus
‘Y’, with ‘X’ and “Y’ for all practical purposes identical in function/meaning.

The requirement ‘X’ = ‘Y’ could be met in two ways, only one of which has so far been
taken into consideration. Either, *sunu- and *suyu- could have been derived from the same
base by means of different, but functionally equivalent, suffixes, or, *su-nu- and *su-yu-
could be compounds with different, but meaningwise equivalent, second elements.

There seems to be reason to think that Proto-Indo-European ‘root nouns’ could refer to
actor and action/result alike. It is then not improbable to assume that OlInd. su- ‘begetter’
was matched by a concomitant ‘act (result) of begetting (giving birth)’. We may thus
tentatively postulate an identification of *su- in *sunu-/suyu- not as *‘genitor’, but as
*'generatio'. In an informal componential semantic analysis, ‘son’ could clearly be identified
as ‘male member of the first descending generation’.

407

If this line of argumentation is plausible, we have to find two Indo-European roots or bases
that could both be used to express the notion ‘first descending’ and both had the shape
*nu- and *yu-, respectively, in the zero grade. In conventional full-grade notation, the roots
would have to be *new- and *yew-. The twofold requirement is met by two items: the first
root mentioned occurs in reconstructed *newos ‘new’ (cf. Pokorny 1959: 769); *yew- is
found in Olnd. ydvista- ‘youngest’ and numerous related forms (cf. Pokorny 1959: 510—
511). I take *sü-nu- and *st-yu- to be exocentric compounds to be paraphrased as ‘one
who belongs to the new (young) generation’; the relationship expressed in exocentric
compounds is taken to be unspecified — as in genitival constructions; possession is viewed
as merely a special case of a more general relationship.

As far as form is concerned, *-nu- in *su-nu- would be the first attestation of a zero-grade
form of *new-. In spite of its wide distribution, PIE *newos must be a fairly recent forma-
tion (with its thematic stem and no trace of ablaut resembling PIE *deywos ‘divine’ beside
PIE *dyew-; cf., e.g., Pokorny 1959: 184—186). Whether *yew- should be left as it is or
rewritten as *Xyew-, cannot immediately be decided. A decision in favor of the second
alternative would permit us to propose a connection with forms such as Olnd. äpu- ‘life’,
Gk. aief *always', etc. (for a list of relevant forms see Pokorny 1959: 17): this connection
seems possible, but not necessarily more than that. There is, however, a further point
which may be mentioned here:

Beside *sunu- we find *sunu- reflected in a number of languages (Germanic, but possibly
also Iranian, though here the Old Indic parallel supports a reading *hunu-). Loss of length
in unaccented position, as advocated by Dybo and Szemerényi, is not a well-supported
explanation. Suppose now that *-yu- in *suyu- was to be derived from *Xyew- and *-nu-
from *new-, and suppose the base to be posited for ‘beget; give birth’ was an anit root
*sew-, the regular results of composition would be *su-nu- and *su-Xyu-, the former then
with a regular short vowel in the first syllable, the latter with what usually (though not in
Tocharian) developed into a long vowel. Suppose further that *sunu- and *suyu- existed
side-by-side in pre-Balto-Slavic and pre-Indo-Iranian: it would then be very easy indeed to
postulate secondary lengthening of *sunu- under the influence of *suyu- with regular
length.

One major issue has not been discussed by now: The order of segments in the alleged
exocentric compound is the reverse of what we normally find: not modifier +head as in
longlegs, but head+modifier. Rather than take this to invalidate the proposal made here, I
would like to suggest that the compounds denoting ‘son’ be included among the relatively
small number of compounds of the head+modifier type (Gk. helkekhitones, helkestpeplos)
which I take to be survivors from a time when Proto-Indo-European was a language of the
VX — and not, as arguable for a later period, of the XV — type. Since I presented indepen-
dent reasons for this claim elsewhere (in Winter 1969: 217—218 and in a lecture given at
the Historical Linguistics Conference held in 1983 at Poznan, the latter to be published
shortly), I shall limit myself to this brief statement, but add to it the further comment that
if the interpretations offered here are acceptable, the inherited terms for ‘son’ in Indo-
European would belong to a very old layer of the reconstructed language, and that this
claim apparently applies to both *sunu- and *suyu-.

408

References

Adams, Douglas Q., 1984. ‘The position of Tocharian among the other Indo-European languages’, JAOS
104: 395-402.

Brugmann, Karl, 1906. Vergleichende Laut-, Stammbildungs- und Flexionslehre nebst Lehre vom Ge-
brauch der Wortformen der indogermanischen Sprachen. (Second edition) 2:1: Allgemeines. Zusam-
mensetzung (Komposita). Nominalstämme (Straßburg: Trübner).

Pokorny, Julius, 1959. Indogermanisches etymologisches Wórterbuch 1 (Bern — München: Francke).

Solta, Georg Renatus, 1960. Die Stellung des Armenischen im Kreise der indogermanischen Sprachen.
Studien zur armenischen Geschichte 9. (Wien: Mechitharisten-Buchdruckerei).

Szemerényi, Oswald, 1977. Studies in the kinship terminology of the Indo-European languages with
special reference to Indian, Iranian, Greek, and Latin. Acta Iranica, Textes et mémoires 7, Varia
1977 (Téhéran — Liège: Bibliothèque Pahlavi; Leiden: Brill).

Van Windekens, Albert Joris, 1976. Le tokharien confronté avec les autres langues indo-européennes 1:
La phonétique et le vocabulaire. Travaux, Centre International de Dialectologie Générale, Université
catholique néerlandaise de Louvain, 11 (Louvain: CIDG).

Walde, Alois — Pokorny, Julius, 1927. Vergleichendes Wörterbuch der indogermanischen Sprachen 2
(Berlin — Leipzig: de Gruyter).

Winter, Werner, 1969. ‘Vocative and imperative’, in: Jean Puhvel (ed.), Substance and structure of lan-
guage (Berkeley — Los Angeles: University of California Press), 205-223.

Winter, Werner, 1985. ‘Tocharian B soy, A se and related forms’, JAOS 105: 259-264.

Inscriptionis palaeo-Osseticae apud Zelencuk flumen repertae lectiones
quaedam novae proponuntur

Archicles Zgusta (University of Illinois)

Graecos, qui ipsi populorum Semiticorum quondam discipuli fuerant, multas alias nationes
gentesque artem scribendi docuisse inter omnes constat; fons tamen scientiae Graecia non
solum temporibus antiquis erat, sed etiam aetate recentiore, quam Byzantinam appellare
solemus.

Gentes Iranicae, quae in ora septemtrionali Ponti Euxini nec non a Caucaso nivoso septem-
trionem versus vivebant, bis cum hominibus, rebus, litteris Graecis commercium habuerunt.
Quas Scythicas Sarmaticasque fuisse scimus gentes iam ante aetatem Christianam cum
oppidis Graecis in litore Ponti Euxini conditis commercium haud minimum habebant:
singulos Scythas Sarmatasque in oppidis Graecis vixisse testimonio et auctorum et inscrip-
tionum novimus nec non res originis Graecae varias, ornamenta, vasa, supellectilem do-
mesticam dico, in Scytharum Sarmatarumque tumulis repertas esse scimus. Quamquam illi
auctores antiqui Graecorumque oppidorum inscriptiones nonnulla vocabula et nominum
complures centurias nobis proferunt, tamen Scythas vel Sarmatas ipsos Graece scribere
potuisse nullo testimonio cognovimus.!

Post multa saecula, Scytharum Sarmatarumque nomine in Ponti Euxini litore septemtrionali
iam propter populorum magnam permutationem, quam gentium migrationem appellamus
(saec. V, VI aet. Christianae), oblivione iam obruto, solum posteri eorum, Alanos dico, in
eis regionibus habitabant; quorum Alanorum Ossetae hodierni progenies sunt. Aetate
nostra Ossetae in medii Caucasi vallibus et promunturiis vivunt; attamen nomina locorum
et imprimis fluminum testimonium tale dant, ut dubitare non possimus, quin Ossetae anti-
qui medio aevo immensam illam regionum planitiem usque ad Maeotidem et ultra Tanaim
ad Danaprim usque extentam tenuerint?

Qui Ossetae antiqui ab imperio Byzantino non solum religionem Christianam receperunt,
sed etiam eisdem temporibus litteras Graecas didicerunt: quod gentium Iranicarum in hac
parte mundi viventium cum Graecis alterum commercium erat. Tertiodecimo saeculo
medio Guilelmus Rubruquius, qui eas partes terrae visitaverat, Alanos, qui Aas (Ossetae)
appellabantur, Christianos esse et litteris Graecis uti enarrat (qua de re cf. Basilium Joannis
filium ab Aba prognatum in opere laudato p. 268 optime disserentem). Attamen posteriore
tempore litterarum Graecarum memoria interiit, ita ut incipiente aetate nostra, saeculo un-
devicesimo, Ossetae litteras partim Georgicas partim Rossicas accipere debuerint; occasione
data etiam Arabicis litteris paululum scribebant et vicesimo saeculo, post magnum bellum
primum et post magnam rerum Rossicarum permutationem, sed non infra Josephi Bes-

410

sarionis filii dictaturam, Ossetae etiam litteris Latinis utebantur. Attamen retro nobis ad
medii aevi res litterasque properandum est.

Unicam? quam novimus inscriptionem palaeo-Osseticam anno 1888 D.M. f. Strukovius, vir
clarissimus, societatis archaeologicae Petropolitanae sodalis praeclarus, in ripa fluminis
Zelentuk in regione quae Kuban appellatur repperit imaginemque ipsius delineavit et eam
professori Petropolitano Pasicrati Millerio, Rossici imperii Academiae sodali celeberrimo de
rebus Osseticis optime merito, edendam tradidit. Paene omnes sequentes editiones ab hoc
exemplari Strukoviano pendent lapide quod sciam obruto. Pasicrates (Rossice Vsevolod)
Millerius acerrimo ingenii sui acumine inscriptionem litteris Graecis saeculi undecimi vel
duodecimi sed lingua Ossetica scriptam esse recognovit eamque interpretatus anno 1893
optime edidit. Post editionem principem Millerianam editio illa, quam Basilius Joannis
filius ab Aba? prognatus (qui Ossetice Abaity? Ivany fyrt Vaso, Rossice Vasilij Ivanovič
Abaev appellatur), Muscoviae professor eminentissimus omniumque qui nunc florent
Ossetarum decus atque delectamentum, anno 1944 et denuo 1949 publici iuris fecit,
maximae est auctoritatis, cum doctissimus ille vir permultas lectiones interpretationesque
novas et eas Millerianis meliores protulerit.

Titulum, cuius lineamentorum imaginem lucis ope factam in figura prima, lector benevole,
invenies, litteris Graecis scriptum post Millerium clarissimum Basilius Joannis filius, vir
eminentissimus, optime et perlegit et interpretatus est: 1 ‘I(noov) 2 X(piord)c. 3 ò
&yoc^ 4 N(QkóÀa- 5 oc. 6 Zayxn- 7 pn povpr 8 X... 9 ..pn q- 10 0vpr 11 Maxa- 12
vap. Ia- 13 kada(p}n 14 yovpr 15 Avna- 16 À. Avana- 17 Xavn g- 18 ovpr 19 Aak.
a- 20m ténp- 21 de. OOXA.

In inscriptionis initio lesus Christus et Sanctus Nicolaus (qui ab antiquis Ossetis valde cole-
batur) invocantur. Sequuntur nomina virorum defunctorum ordine Ossetico exhibita: in
lingua Ossetica imprimis patris nomen in genitivo laudatur, quod vox furt = fyrt, ‘filius’
sequitur, deinde filii nomen in nominativo. Si exempli gratia editoris clarissimi nomen
perscrutamur, videmus supra eum Ivan-y fyrt Vaso = Joannis filium Basilium appellari;
genitivus in -y desinere solet. Eodem modo viri defuncti in titulo nostro appellantur:
Zaxnpn, ie. (per itacismum aetatis Byzantinae) Sachir-i (gen.) qovor? (7 filius) X... (no-
men filii non legitur). In linea 9 habemus -p1, i.e. -r-i (nomen patris in genitivo non legitur)
poupt Ilaxadap (i.e. Oss. Bagatér). Sequitur in linea 12—13 Ilakada(p)n? povpr Avraà
(ie. Oss. Ambal); 16 Axa)ma\arn!® (i.e. Ambalin-y) povpr Aak (i.e. Oss. Lég). Etiam
sequentia vocabula Basilius Joannis filius optime interpretatus est: avn tSnpde = Oss. ani
cirtà ‘eorum monumentum?.

Quattuor ultimae litterae OOCA neque a Pasicrate Millerio eminentissimo neque a Basilio
Joannis filio doctissimo explicatae sunt. Attamen Gregorius Theodori filius TurCaninovius
Leninopolitanus, philologus celeberrimus, primam litterarum ut solis symbolum O = ©,
secundam ut 9 ' - 9, tertiam ut o '- 200, quartam ut A ' 7 30 interpretatus est; quorum
symbolorum numerorumque significatio secundum doctum Turcaninovium ita intellegenda
est:! symbolum solis O annos per aeram ut dicimus solarem vel per solis revolutiones
computandos esse indicat; oA '(7 230) numerum talium solis revolutionum indicat, quarum
una duodetriginta (28) annos comprehendit: itaque ducenties tricies duodetriginta faciunt
sex milia annorum et quadringentos quadraginta (230 x 28 = 6440); 9 '(7 9) novem annos
in ducentesima tricesima prima solis revolutione iam elapsos esse significat. Hoc modo
annorum numerus sex milia quadringenti undequinquaginta (6449) est, quos Turéaninovius

411

doctus ut dicitur a creatione mundi computatos esse putat; cum primus aetatis Christianae
annus quinquies millesimus quingentesimus octavus a creatione mundi sit, titulum anno
aetatis Christianae uno et quadragesimo et nongentesimo (6449 — 5508 = 941) scriptum
esse mathematicus noster contendit.

Quam opinionem plus ingenii quam veritatis praebere neminem fallere potest: insoliti
litterarum significatus (O = solis symbolum), insolentes numerorum computationes binae,
immo ternae (numeri primum in duas turmas digerendi, deinde secunda turma multiplican-
da, tum omnia addenda sunt), inaudita in inscriptionibus secundum solis revolutiones
annorum numeratio, haec omnia faciunt, ut Turčaninovii docti opinionem veri dissimilli-
mam esse haud sit quin ignoret.

Quae cum ita sint, post editionem Basilii Joannis filii ab Aba prognati duo loci soli in titulo
restant, qui non comprehendantur: lectionem linearum octavae et nonae et ultimarum
lineae unius et vicesimae litterarum significationem dico.

Quod ad primum locum incertum pertinet, me iudice ut sequitur legendus est: Quin prima
octavae lineae littera X sit, non est, qui dubitet; sequitur O imperfectum (dextra circuli
parte sola extante) et clarum B; littera sigma C sequitur, minor quam ceterae litterae: ad
quam rem confer omicron parvulum, primam quintae et primam decimae lineae litteram;
ultima deinde lineae octavae littera H, eta male scriptum esse videtur, ad cuius formam
tertiam lineae nonae litteram conferre velis, quae quin etiam H sit, non est, qui dubitet. °
Prima nonae lineae littera quae dicitur ligatura oro esse videtur,!^ cuius omicron imperfecte
aut scriptum aut delineatum est; ligaturae lectio ore veri minus similis est. Sequentes litterae
PH® bene leguntur. Quae supra de nominum Osseticorum ordine diximus in mente haben-
tes, videmus post furt ‘filius’ (lin. 7) in linea octava nomen filii in nominativo sequi debere;
nec minus necesse esse, ut alteri vocabulo furt (lin. 9, 10) nomen patris in genitivo praece-
dat: quod patronymicum in -pn (= -ri) cadere iam constat. Quae cum ita sint, lineas sextam
usque ad duodecimam duobus modis legere possumus: aut Zaxnpm co[u](op7) XoB, Znoropn
youpT llaka2ap aut Zaxnpm vo[vl(pr) XoBs, Horopn etc.; itaque patris Sachir appellati
filius aut Chow aut Chows et Baqatari pater aut Sistor aut /stor appellabatur. Lectio sive ut
ita dicam verborum delimitatio Xofic, Horopr; multo melius quam Xofl, Enoropn est, quia
nomen Horop optime explicatur: osseticum stur *magnus' cum vocali prothetica [i-].!^?

Quod ad ultimas unius et vicesimae lineae litteras pertinet, piam formulam Christianam
prae oculis habere credimus. Litteras OC plane vocem O(eóy breviter scriptam esse fallit
neminem; ultima littera ut nostra fert opinio non A sed litterae N imperfecta aut scriptura
aut delineatio est, ipso N pro Graeca voce vika "vincit? breviter scripto stante (Turyn 1972:
tabula 49). Cum titulus noster Christianus sit, optime in initio invocatio 'I(nooU)c X(ptoróyc
et in fine formula ó G(eóy [v](uà) ‘Deus vincit’ legitur, quae formula in omnibus aetatis
Byzantinae scriptis frequentissime reperitur.

Itaque secundum nostram opinionem inscriptio hoc modo legenda vertendaque est:

1 ‘I(noovs) 2 X(pwTd)s.

3 Oayuocs) 4 N(i)KdAa- 5 oc.

6 Zaxn- 7 pn golu](or) 8 Xofc.

H- 9 oropn y- 10 ovpr 11 Ilaka- 12 dap.
Ila- 13 xada(p)m 14 govor 15 Avra- 16.
Avra)ma- 17 Xavn y- 18 ovpr 19 Aax.

a- 20 vn Tinp.- 21 de.

ó O(ed)c [r](Kd).

412

Versio: Iesus Christus

Sanctus Nicolaus

Sachir-ii!° filius Chovs
Istor-ii filius Pakathar.
Pakathar-ii filius Anpal
Anpalan-i filius Lak.

Eorum monumentum (stela).
Deus vincit.

Media inscriptionis pars hodierna lingua Ossetica reddita et litteris Latinis expressa talem
formam habet:

Sachiry fyrt Chovs.
Ystury fyrt Bägätar.
Bägätary fyrt Ambal.
Ambalany fyrt Lag.
Ani cirtà.

Videmus Osseticam linguam saeculi undecimi vel duodecimi iam talem esse, ut eam hodie
novimus: pristinae formae Iranicae (furt, Av. pudra-, Scrt. putra- ‘filius’, stur, Av. stura-
*magnus, fortis) cum formis Caucasicis iam mixtae sunt; imprimis nominum propriorum
syntaxis, quae in Scythicis Sarmaticisque tertii et quarti saeculi nominibus (Zgusta 1954)
aliarum Indoeuropearum linguarum modum (ie. primum nomen filii in nominativo,
deinde nomen patris in genitivo) sequebatur, in inscriptione nostra iam formam hodiernam,
ab origine extraneam habet.

Quae res maximi momenti est, quippe quae Osseticam linguam iam ante saeculum undeci-
mum vel duodecimum formis modisque Caucasicis infusam esse demonstret.!9

Notulae

De omnibus his rebus viri docti Max Vasmer (1923); V. I. Abaev (1949); J. Harmatta (1950-51:
261—312; L. Zgusta (1954), aliique bene egerunt. Cf. etiam notam 3.

De quibus rebus in Abaevi doctissimi opere supra laudato lector benevolus plus legere poterit.

Alias inscriptiones seriticis litteris scriptas, quae difficillimae interpretationis sunt, palaeo-Osseticas
esse contendit G.Th. f. TurCaninovius (1971).

V. Miller 1893:110—118.

Aba (et Latine et Ossetice) nomen totius gentis progenitoris prisci est; Oss. Abaity nomen gentis in
genitivo pluralis est: *Abaetidarum".

Abaev 1944; denuo typis expressum in Abaev 1949:260—270. In hac editione etiam tituli exemplar
lucis ope depictum est, quod in opusculum nostrum recepimus.

Lectio à &yws non est satis clara. Pasicrates Miller, Osseticarum rerum indagator primus, oa.« legens
tertiam litteram ut Graecum 7 interpretari voluit, in vocabulo oarc Osseticum wac ‘sanctus’ inesse
subodorans; quae lectio valde dubia est, imprimis quia tertia littera Graecum 7 esse non potest,
deinde quia Sanctus Nicolaus Ossetice wac Nikkola appellatur: Osseticum oar« cum Graeca nominis
terminatione -oç minime congruit, ut optime Basilius Joannis filius ab Aba prognatus perspicacissi-
mus cognovit. Cuius viri doctissimi lectio 5 ttyvoc multo melior est: tertia ligatura fortasse ut yi nimis
cursim scriptum legi potest; ultima autem littera, circulus ille medius, omicron cum littera sigma
ligatum et imperfecte vel scriptum vel delineatum est. Licet lectio incerta sit, nihilo minus vocabu-
lum à tt'yoc uno vel altero modo legendum est.

Exemplaris imaginem perscrutatus vocem qovpr completam non esse, sed stelae dextrum marginem
locum litteris supplendis praebere non posse vides, lector optime: itaque, si exemplar recte delinea-
tum est, wo|voz] supplere non possumus, sed errore lapicidae assumpto aut go(up)[7] aut yofu (v7),
quod veri similius esse videtur, emendandum nobis est.

413

1 IE

ZA "E

3 DA,..2

4 NIKO AA
o OX

6 XAXH
7 PHOOY[PT]

9 ...PHo
10 OYPT
11 TIAKA
12 OAPTIA
13 KA®AN
14 DOYPT
15 ANFIA
16 AANATIA
17 AANH®
18 OYPT
19 AAKA
20 NHTZHP
21 OE-00XA

414

(p) emendandum, non supplendum est, exemplari nullum locum praebente.

Secundum a delendum esse videtur: (a).

Cf. Turtaninovii docti epistulam ad Basilium Joannis filium scriptam atque in commentario ad suam
editionem (1971:267 sq.) ab eodem viro docto publici iuris factam.

Has ultimae lineae quattuor litteras Turtaninovius doctissimus (1971:105) eodem modo in libro suo
(de quo videas supra in notula 3) interpretatur, Rossici tamen textus, qui ‘chronologicus’ a recentio-
ribus editoribus appellatur, brevem adumbrationem addens. Saeculo duodecimo Cyricus monachus in
Rossicis litteris non ignotus (Rossice Kirik appellatus), monasterii Antoniani in Novgorod urbe dia-
conus et domesticus, de quae chronologia dicitur egit; quae pertractatio in virorum clarissimorum S.
P. f. Obnorskii et S.G.f. Barchudarovii libro (1938:186 sqq.) doctiore publici iuris facta est. Mona-
chus doctus certiores nos facit se anno ‘post Adam’ sexies millesimo sescentesimo quadragesimo
quarto (6644) scribere, quod secundum alium Cyrici locum, in quo de revolutionibus solis agitur, du-
centae triginta septem solis revolutiones totae et octo ducentesimae duodequadragesimae revolutio-
nis anni sunt (blo ot adama s’I’ z’ a poslédneg kruga idetb osmoe léto; lin. 117 sqq.). Ducenties
tricies septies duodetriginta faciunt sex milia sescentos triginta sex annos (237 x 28 = 6636); additis
octo ultimae revolutionis annis eundem numerum annorum, 6644, obtinemus. Haec Cyrici computa-
tio totius interpretationis vel ratiocinationis Turtaninovianae solum fundamentum est. Attamen Cy-
ricus, diaconus domesticusque doctus, non de annorum in rebus gestis indicandorum modo sed de
aliis rebus partim astronomicis, partim astrologicis hermeticisque, partim ad religionem pertinentibus
agit; exempli gratia dico učenie imYe védati Elku Eisla vséch [$t (1.38: quo modo numerus omnium
annorum statuendus sit’), kako es(tb) razuméti Cislu dnevnomu (1.62: quo modo numerus diei in-
tellegendus sit’), utenie o in)dikté (1.86: "indictionis doctrina), kako mofeth védati sintny krug
(1.103: ‘quo modo revolutio solis intellegenda sit’), sice Ze razum&ti mozet krug lunny (1.121: ‘hoc
modo revolutio lunae intellegenda est’), o věcech mira (1.139: ‘de mundi aetatibus’), o pronovlenii
nbs (1.143: ‘de caelorum renovatione’), na kolicé lété obnovleetse more (1.152: ‘in quantis annis
mare renovetur’), alia id genus. Computatio annorum per solis revolutiones, annorum turmas in-
dictionis instar recurrentes, non ad res gestas applicatur; Cyricus ipse, monachus eruditus, cum de
rebus gestis annalium modo agit, numerum annorum nec per solis nec per lunae nec per alias revolu-
tiones indicat. Cf. lin. 1 sqq.: Ot adama do potopa, v’s’ m’iv’ léta. ot potopa do razmésénia ezykb
f U, etc. 'Ab Adam usque ad diluvium duo milia ducenti quadraginta duo anni; a diluvio usque ad
linguarum confusionem quingenti triginta, etc.'. Itaque Cyrici textus chronologicus dictus ad quat-
tuor ultimas inscriptionis nostrae litteras interpretandas minime pertinere videtur.

Etiam alias litterarum formas minus recte ad normam et minus constanter scriptas conferre potes:
N in linea quarta; P in linea septima (cuius tamen quod hasta transversaria esse videtur solum exem-
plaris pro lapidis rima error plane est) a forma eiusdem litterae in linea quarta decima nec non duo-
devicesima valde differens; Z, quartam vicesimae lineae litteram; litterae H formas inconstantes in
linea sexta, septima, nona.

Cf. Gardthausen tabulam 7, columnam ad annum 1059 pertinentem.

a Opusculo nostro iam perfecto Basilius Ioannis filius ab Aba prognatus per epistulam pridie Kal. Mart.
anni aetate Christ. 1977 scriptam certiores nos fecit G.Th.f. Tur£aninovium celeberrimum anno
1974 de inscriptione nostra disputantem lineas octavam et nonam Xoflc Iorop perlegisse eaque voca-
bula *magnus Ows' interpretatum esse, quia owsi in lingua palaeo-Georgica Ossetarum nomen pris-
cum fuerat. (Turtaninovii clarissimi disputatio numquam publici iuris facta est.) Vocem ‘magnus’ in
secundo vocabulo bene recognovit Turéaninovius doctissimus; tamen Horoprm non ad nomen praece-
dens pertinet, sed nominis sequentis IIakadap patronymicum est. Si in linea octava cum Turcanino-
vio doctissimo nomen Of legere volumus, quid sibi prima eiusdem lineae littera X velit, nescimus.
Quam litteram si T legere possemus, bene esset, quia ad vocem praecedentem pertinere posset:
HOLIT = eo[v](o)l r; quae lectio optima esset. Attamen haec littera (octavae lineae prima) solum X,
non T esse potest, si lineamenta Strukoviana bona sunt; cf. X, tertiam sextae lineae litteram, et T
quater in lineis decima, quarta decima, duodevicesima, vicesimaque bene scriptum.

-į latina genitivi terminatio est.

Praesente opusculo iam scripto Iohanna Nichols, femina doctissima et rerum Caucasicarum peritissi-
ma, ad ea, quae la. S. f. Vagapov (1980:100--117) publici iuris fecerat, animum meum advertit.
Lectiones alteras paene nullas Vagapov doctus praebet, sed titulum nostrum lingua Ingussica inter-
pretari vult. Qua de interpretatione in altero loco loqui volumus; satis sint observationes tres. Primo:
Vagapovii docti lectio Sachiri fu(rto)j: Chob(a) Sitleri furt, Bagatar Bagatai furt, Anbalan(a) Abalani

415

furt (i.e. Latine "Sach filii: Choba Sitler-ii filius, Bagatar Bagata-ii filius, Anbalan(a) Abalan-ü
filius") nullam interpretationem praebet: si Choba filius Sitleri, Bagatar filius Bagatai, Anbalana filius
Abalani est, qui sunt Sachiri filii? Deinde: Quartam vicesimae lineae litteram epsilon esse (sicut vellet
Vagapov doctus) haud est, qui credat: velit lector secundam vicesimae primae lineae litteram com-
parare, quae verum epsilon est. Postremo: Non est, qui credere possit, quattuor ultimas vicesimae
primae lineae non solum sinistrorsum legendas esse, sed etiam propaenultimam earum pro A, paenul-
timam pro À fungi. — Opusculum hoc ante iam multos annos scripsi, sed propter fatorum invidiam
Hoenigswaldii nostri doctissimi carissimique volumen gratulatorium non sine cunctatione ac mora
publici iuris factum est. In quibus annis haud pauca ad nostram inscriptionem spectans materia col-
lecta est: quam in opere The Old Ossetic Inscription from the River Zelentuk ab Academia Vindo-
bonensi edendo plenius tractabo.

Operum adhibitorum conspectus

Abaev, Vasilij Ivanovit, 1944. “Drevneosetinskaja zelentukskaja nadpis’” [The Old Ossetic inscription of
Zelenchuk |, Soobenija Gruzinskoj Akademii Nauk 5.2:217 —226.

Abaev, Vasilij Ivanovit, 1949. Osetinskij jazyk i fol’klor 1 [Ossetic language and folklore 1], (Moskva-
Leningrad: Jzdatel’stvo Akademii Nauk SSSR). (Contains reprint of Abaev 1944 on pp. 260—270).

Gardthausen, Viktor Emil, 1911—1913. Griechische Palaeographie (2nd edition) (Leipzig: Veit).

Harmatta, Janos, 1950—1951. “Studies in the language of the Iranian tribes in South Russia", Acta
Orientalia Hungarica 1: 261—312.

Miller, Vsevolod F., 1893. “Drevne-osetinskij pamjatnik iz Kubanskoj Oblasti” [An Old Ossetic monu-
ment from the Kuban region ], Materialy po archeologii Kavkaza 3: 110-118.

Obnorskij, Sergej Petrovit/Barchudarov, S.G., 1938 (1952). Chrestomatija po istorii russkogo jazyka 1
[Chrestomathy on the history of Russian] (Moskva).

Turčaninov, Grigorij Fedorovit, 1971. Pamjatniki pisma i jazyka narodov Kavkaza i vostočnoj Evropy
[Monuments of script and language of the people of the Caucasus and Eastern Europe] (Leningrad:
Nauka).

Turyn, Alexander, 1972. Dated Greek manuscripts 2 (Urbana: University of Illinois).

Vagapov, J.S., 1980. “O jazyke zelenéukskoj nadpisi” [On the language of the Zelenchuk inscription],
Voprosy vajnachskoj leksiki (Groznyj: IIOF Ce$eno-InguSskoj ASSR), 100-117.

Vasmer, Max, 1923. Untersuchungen über die ältesten Wohnsitze der Slaven 1. Die Iranier in Südrußland
(Leipzig: Peters).

Zgusta, Ladislav, 1954. Die Personennamen griechischer Städte der nördlichen Schwarzmeerküste (Prag:
Nakladatelstvi Ceskoslovensk& Akademie V&d).

A KM Laryngeal as a conditioning factor for s-loss in Sora-Juray-Gorum

Arlene R.K. Zide (Loop College, C.C.C., and University of Chicago) and Norman H. Zide
(University of Chicago)

1. The Austroasiatic phylum of which the Munda family is the westernmost! branch is
notable for the complexity of its vowel systems. The Munda languages have less elaborate
vowel systems than do their Mon-Khmer relations, but present numerous problems in
reconstruction of their vocalism. One of these problems is the unexplained loss of s in
Sora-Juray-Gorum in a small number of old core vocabulary items.

In this paper, we offer one explanation for this loss of s in otherwise ‘inexplicable’ circum-
stances in several Munda languages, notably in Sora-Juray-Gorum? and independently in
Gorum.

S-oss? at the SJG stage is exemplified by a small set of cognate pairs showing ( in SJG and
s in GRG (and elsewhere in Munda). Most of these forms have been known for some time
(see Pinnow)^. Like a number of other sound changes in Munda that leave small, residual
sets of problematic cognates (e.g., the forms with low tone in Korku), these forms presum-
ably reflect a once general feature lost in most of the languages and preserved in a limited
set of forms in one or two. We suggest that s-loss in SJG may be conditioned by creak
(V) or creaky vowel — more precisely, s is lost before creaky voice. The creak hypothesis
is new — nothing like it to account for s-loss in SJG has been suggested previously.

Table 1
GRG SIG Other Munda KM English
1. — — *Vren Kharia sorey *sVren ‘stone’
2. *-si SJG *i, *si Jouse”
(h)i, (h)i?
SI *i?i/-i
Go. agid/-id
3. *sul- *ol- *sol- ‘to crow’
4. *son- *on- *s91)- ‘to sell, measure out’
5. *-sob- *9- *s9 ‘winnowing
S. oj-yer/-a basket?
J. ojfoHyer
Go. oh-yer
6. *sVnol *anal *sVial” ‘firewood, fuel”
7. *sVñ(lob SJ *ane2b, *sVn(1)ob ‘tree, bush”
Anob
8. *sVlu2b *Ale2b, alob *sVlob *chevrotain, antelope'

9. *si(n)ron Go. in(=)ran *sinran ‘medicine’

418

GRG SIG Other Munda KM English
10. Re. suti? SJ SA" *sVti ‘female genitalia’
?GRG *sV(blogc'
11. sVbkjen- So. ufiul-da *sVlo/uñ- ‘to perspire’
Gu. solen-
Re. suleg-
Gt. solæ-
12. *sa()I- *(Vn)A?]/AHI Kh. sol *sal ‘mortar, husking hole”
Gt. sa(-)bo? So. anal Ju. esod
Re. sa(-)bu J. onalad Mu. sgel
Go. a?al
13. *sam- Go. am-sad- *sam-? ‘remove a pot
Re. sam- from the fire'
14. *se2r *ar- NM *sVreXn-1° *SEI- *have a hoarse
GR *ser- So. argo- [ser] throat, choke,
“choke, suffocate” sing, shout’
Go. aH-ran-
sid.”
arur-
“be hoarse”
aH-rig-

“choke on a bone”

2. Creak in present-day Munda Languages

V (creaky vowel) is found phonemically in Gorum, and phonetically in Juray.!! No creak
has been recorded for any of the other Munda languages nor, to our knowledge, for any
other language of the Indian area. Insights into creak can be seen in the cognate phenomena
(e.g., of Gorum creak with GRG). Some cognates of forms with creak in Gorum (and with
reconstructed creak in SJG) have unexpected consonantal reflexes in GRG (e.g., Go.
toroH-, SJG *toro ‘to snore’, Gu. torod-, Ga. torhwa?-, GRG *tVr(h})od- (?); Go. bioH-gi,
SJG *biysH; Gu. bier ‘yesterday’. Again the number of relevant cognates is small. Other
examples of Gorum VH (ie., V) include: teHmi, SJG *teHme, tej2-me ‘new’, GRG *tVme
(Gt. tmi, Gu. time); Go. jeHmol, So. jab-mol, SIG *jeHmol, jej2-mol!? ‘seed, paddy seed’,
GRG *cV( )mol(?), Gu. sugmol, Re. sumu, Gt. cmu.

In Juray, creak seems to be primarily a function of the breaking of the vowel /e/ into a
diphthong [29]. However, there is a tiny residue of forms in which, phonetically, one gets
[V] which cannot be attributed to this breaking. The forms are: re(H)(-Jsa-s£j ‘infant,
newborn’ (contrast re( -/sa ‘kind of neem tree used for hookworm"); guHgud- ‘to scratch
oneself’ (cf. Gorum guguH- and Sora ged- with the same meanings); fega(H/ ‘friends and
relatives’; and (d/ri(H]yab-mad- ‘to blink’; also to be considered possibly is siH/?-siy-gal-
‘to whistle’.

While there are a number of other instances of creak (written here as 7) in Juray, almost all
other examples could be explained as a) either dissimilation of a consonant before an
identical consonant, especially of -d’ before d-, or b) as is probably the case even for
siH /?-sig-gal-, a kind of weakening (or fricativization?) of glottal stop.

419

A residue of forms is, however, not susceptible to this explanation. guHgud- might have
been attributed to an ordinary reduplication of the form gud- ‘to scratch’ except that the
cognate forms as well as the morphophonemics are suspect, and would indicate a more
complex provenance. ? The extremely small number of forms with H (V) and the fact that
H only appears in one dialect of Juray, however consistently, rather than another (hence
the parentheses around H in most cases) leads us to treat creak in Juray as a phonetic
rather than a phonemic entity.

In Gorum, however, H is comparatively common, with both a wide enough distribution as
well as essentially phonemically contrasting environments to warrant treating it as a pho-
neme. Cf., for example, aHdun- ‘to screen, obstruct’ and a?dor- 'to trouble, bother, pester'.
Af occurs in a variety of positions, except initially. Thus, one gets /aHsuy/, CF /-suy ‘house’;
/goHsay/- CF /-say ‘cooked rice’; /oH-der-/ ‘to winnow grain’; /denH/ ‘contingent affix,
affected’; /boHb-re/ ‘sibling’, /guguH-/ ‘to scratch oneself etc. Unlike the glottalized con-
sonants, there seems to be no feature of release associated with it preceding vowels. H oc-
curs primarily as marker of the [affected] function of the subject NP on the verb on
which it is affixed. Creaky voice in Gorum, unlike the glottalization of obstruents, must be
represented in underlying forms — there being no way to predict its occurrence or non-oc-
currence. Further, we do not get opposable sets of +/-creaky vowels; nor is H regularly
derivable in any obvious way from glottal stop or from any other segments in the language,
although some H are certainly derivable as a secondary weakening of glottal stop.

In fact, the nature of creaky voice in Gorum (and Juray for that matter) deserves some
closer attention. There is a close relationship which obtains between the glottal stop and
what may be tentatively labelled a laryngeal fricative.'* If any simple characterization can
be offered for it, the laryngeal (or creak) in Gorum is an intermittently voiced fricative
homorganic with glottal stop. If (following Ladefoged and Catford) we understand creaky
voice as a point on a continuum — i.e., in terms of a glottal parameter — then it appears to
be essentially more akin to a ‘feature’ than to a segment. In Gorum, however, creaky voice
is not just a ‘bound’ feature added onto other segments, or comprising a part of such seg-
ments (e.g. as is voicelessness, or release), but rather, its behavior indicates an independent
segment rather like glottal stop. What the historical situation that the reflex ‘creaky voice’
represents is, however, a different question.

In any case, an investigation of creaky voice (in Juray) has been made using the spectro-
graph.

As in Gorum, [H] is utilized as the symbol for creaky voice, [V], in Juray; in one of the
three idiolects under study it was the obligatory substitute for the glottal stop. It is found
also (optionally) as a substitute for [y?] and sometimes [d?] before another consonant
(especially if an obstruent) before morpheme boundary.

In spectrograms, creaky voice has a characteristic configuration and can be readily iden-
tified. It looks like a fricative, but shows up at lower frequencies than fricatives generally
do; there is a slowdown of beats of the larynx; there seems to be less acoustic energy and
this shows up in the spectrograms as interruptions or gaps in all the formants.! It loses fre-
quency (i.e., presumably goes lower), and where it is the substitute for glottal stop /?/ and
appears between two like vowels (primarily in the idiolect of B. Boya), it shows a 10—15
millisecond gap between vowels, with each closure of the glottis showing up as a line of
burst.

420

The spectrograms of Juray H thus confirm the acoustic phonetic impressions one has of
creaky voice in both Gorum and Juray as a ‘fricativization’ of glottal stop, or an inter-
mittently vibrated ‘glottal stop’.

3. The Cognate Phonema of Creak in GRG

GRG reflexes of KM creak (as reconstructed from Gorum H, and SJG s-loss) have been
identified in the consonantism, the vocalism, sometimes in both the consonantism and
vocalism and in a few cases we find no trace of a reflex of creak at all.

GRG consonantal reflexes of creak can be grouped into those with non-glottalized con-
sonantal reflexes and those with glottalized consonantal reflexes. The first set involves a
continuant somewhere in the word (either a nasal or r or /). The second set includes the
consonants b, d, g (j), ? and h (the first four are weakly voiced or devoiced pre-glottalized
stops).

Let us consider the set involving some sort of glottalic or laryngeal involvement:

b: GRG *-sob, Go. oH, So. 9j/-2, SIG * ‘winnowing fan’; Gu. tab(-/mi-, Go. taH(-)/mu-nuH
‘sneeze’.

d: GRG *tVr(h)od- (Gu. torod-, Gt. torhwa?-); Go. toroH-, So. tara- ‘snore’; GRG *gud- ‘to
comb, pull, sweep’; SIG *gud- but see note 13 regarding Go. guguH-, So. ged-, Ju.
guHgud-.

g: Gu. irig CF-rig; Go. ariH ‘millet, Mandia (Eleusine indica)’; GRG *blj-tir- with derived
noun b-in-i?-tir and verb big-tir in Gutob, ‘spittle, spit’ (Re. buitur, Gt. bwe?tur) Go.
bi?/j-tur- ‘spit’ but beH-rol-nuH ‘to spit up (of babies)”.!9

j: cf. big-/bi? above, and probably irig, GRG g after high front vowels is from earlier (pre-
GRG ?)j.

?: Gu. a?( - dur-, Go. aH-dur- ‘to jump across’. Note that Bhattacharya (for Plains Bonda
(Remo)) has su? as well as su- for *to crow' (GRG *sol-; possibly *sol-). Another inter-
esting but problematic form that is relevant here is the Gt. compound verb wa°-sar- ‘to
call by shouting out loudly’ made up of wa- (without glottal stop elsewhere) ‘to call’
(GRG *wai-) and sar- ‘to sing, sing out’ (GRG *se2r-), the cognate for the latter in SIG
undergoing s-loss. We suggest that the ? in wa? reflects creaky voice in the morpheme,
*se2r-, which follows.

h: The only KM language that has a fully phonemic h is Gta?. Gt. torhwa?- (GRG
*tVr(h)od-) S. tor2-, Go. toroH- (SIG *toroH-) ‘to snore'.

The second set, involving what can loosely be termed continuants, is examined next:

N: Gu. taNdar, So. tadar, Go. taHdar, ta?dar ‘rooster, fowl’: One could consider the Gutob
form to have no reflex of creaky voice, but to liave (as would similar forms without
histories of creaky voice) *spontaneous nasalization' beforc voiced stops. This is plausible
enough, but for the other example of N, this explanation is not available. GRG *cVmol,
SJG *jeHmol ‘seed, paddy seed’. The Gutob is sugmol (Remo sumu, Gt. cmu). One
would not expect a nasal before the m otherwise; creaky voice, we suggest, was reflected
somehow, and the back segment or feature was (then) nasalized (in assimilation to the
following zt only in Gutob). Thus, GRG *cVmol.

421

rt: Gu. bier, Re. er-og ‘tomorrow’; So. biyaji, Go. biyoHgi ‘tomorrow’, J. biya ‘some day’.
It looks as if the GR word for ‘tomorrow’ was composed of two pieces, the second
being -tog (we find -tog elsewhere, as in Gu. mar-tog ‘the day after tomorrow’, mun-tog
‘two days after tomorrow’,etc. Thus the r may reflect not simply V but V plus a follow-
ing -1).!”

1: This example is problematic, and may involve a verbal morpheme with creaky voice and
not the stem ‘to come’. Gu. olo (defective imperative) ‘come!’, perhaps from ol-a (with
-a the intransitive imperative marker), Go. 2H-ay ‘to come’. (Note too Korku suppletive
ol-en ‘went’.)

Vocalism: Our observations on vocalism are more tentative since the regular KM vowel
correspondences have not been as solidly reconstructed as the consonant correspondences.
The forms listed here are those where the GRG and SJG correspondences are notably
deviant.

GR wi- (ui-), Gt. (Plains) we-, Gt. (River) wai-; Go. uiH(-nuH) ‘to go'?, GRG *tVrhod-,
SJG *toroH- ‘snore’, and Gu. bier, Re. er-og, Go. biyoHgi etc. as mentioned above.'?

Those forms showing no obvious traces of creaky voice in GRG are:

GRG *tVme, Go. teHmi, So. tab-me ‘new’, (SIG *teH-me, SJ *tej2-me)*°, GRG *go- in
Re. go-lig- ‘to drape, wear a garment (of women)’, and go-si- ‘to drape/wear a garment (of
men)’, Go. goHtuy ‘cloth, skirt’, So. gad-tuy, J. gad-tuy, gaH-tuy.*1 GRG *bobrefi, bobrefi
‘brother (female speaker)’, So. bab-re, Go. boHb-re ‘id.’ GRG *-ne ‘past participle marker’,
(SJG ??*boHb-re) Go. -nuH ‘infinitival marker, affected’.

With regard to *-no/*-nuH and possible morphological affixal H morpheme(s) N. Zide
noted (1960) that low tone/aspiration in Korku was significantly more frequent in verbal
desinences, and nominal genitive and accusative-dative than one would expect merely by
chance. Something similar has been observed for Gta? with respect to glottalized vowels
(associated with weak stress) in verbal desinences and the genitive. It was also noted (N.
Zide 1973) with respect to the reconstructed PM numerals that it appears as though one
should reconstruct a ‘morpheme of glottalization’ for all or most of these. Are there, then,
affixal morphemes represented (or substantially represented) by H ? Is H (also) a junctural
marker indicating boundaries between, say, certain classes of stem and affix morphemes?
Further, are there ablaut-like phenomena resulting from infixed creaky voice as a morpheme
(or morphemes)? The data invite us to suspect, but not as yet demonstrate, all three.

4. The S-loss Hypothesis

Earlier hypotheses regarding s-loss had taken account of distinctive vocalism, accentual
. phenomena? and glottalization as conditioning factors. However, with limited success.
Glottalization, i.e., both glottal stops and consonants, and N. Zide’s ‘glottalized vowels'
did not solve the problem, although this seemed (and seems) more relevant than say,
accentual phenomena.

The examination of creak in this context arose from a consideration of the possible pho-
netic properties of the conditioning factor. Creak, or creaky voice can be interpreted as a
spirantal feature in view of the phonetic (including spectrographic) data as discussed earlier.
One motivating condition for s-loss might be a constraint on multiple spirants, which
results in the despirantization of the first spirant in a word.

422

The only spirant we reconstruct so far for Proto-Munda is *s.? There are examples in
Korku (see below) of s de-aspirating before a spirantal kh (i.e., [x], e.g., as *sakhàn >
cakhan ‘firewood’. In terms of the syllabic patterns of the Munda languages, consonant
clusters that might spirantize to f, h etc. (e.g. Korku p becoming @/.th in apthae (from
*ab-tha(-)e?) *to wash the hands and feet',^^ or Go. p in apay-u becoming [afay] in allegro
speech) are not found and we see no evidence for reconstructing them. In some KM lan-
guages, there is affrication of what are written [c] and [j] in several of the languages but are
clearly (and reconstructed in the SJG) pre-palatal or palatal stops.

Since Gorum creak looks to be old (it is really odd, and presumably owes nothing to bor-
rowing), we hypothesize that creak and s are the two spirants we reconstruct for KM (and
for PM). We suggest that there is a constraint on spirants in SJG, that is, where s occurs
initially followed by creak later in the morpheme, s is weakened and later lost.

4.1. Parallels in Korku:

Treating s-loss as a reduction of spirants invites consideration of parallels in the modern
Munda languages. There are parallels to such a constraint in Korku, for example.

The development of a constraint on multiple spirants could provide some explanatory
value to the ad hoc accounts given earlier (and the historical reconstruction based on
them). Taking creak as a spirant, and s (these being the only two KM (and PM?) spi-
rants?°) we can state a rule that where two spirants are found in a word the first is des-
pirantized (i.e., s > h, and h later > $). There are parallels in Korku where a sequence of
two spirants in successive syllables is reduced to one — the medial one. (We alluded to this
here above.)

There is some evidence of similar spirant reduction elsewhere in Munda. The two cases are:
Korku s.. $7» c.. s, and (?) kh..x >k..x. For the first case one might expect that an
exception would be made with reduplicative formations — i.e., spirant reduction would not
occur in reduplicatives. (Reduplication is common in the Munda languages, in nouns as well
as verbs.) However, one of the instances of spirant reduction (by one interpretation) in the
Ghatang dialect of Korku occurs precisely with reduplicated infinitives. In these, sod ‘to
fall, drop’, for example, is not — as elsewhere in Korku (and similarly elsewhere in NM) —
reduplicated **so-sod, but co-sod. Likewise for ‘turmeric’ ca-say (elsewhere sa-say). This,
however, must be considered first as one instance of the application of a reduction of
aspirates rule (the sort of rule common elsewhere, e.g., in IA, Greek, etc.) of the following
kind: where a (voiceless) aspirate occurs initially in a monosyllabic verbstem, the initial
aspirate of the reduplicated stem is deaspirated, e.g. kKhad- ‘to enlarge’, ka-khad |kaxäd].
The c in Ghatang Korku is functionally equivalent to an unaspirated s (as k is to kh). The
kh before a and o (depending on the final consonant of the monosyllabic verbstem) tends
to spirantize to x.

In the second case, a few morphemes of *sVkhV(C) (sVkàC) shape are found, e.g., NM
*sakàn ‘firewood’, Ko. sakan or cakan, where the s despirantizes to c (in most, if not all
Korku dialects). For instance, cakhan [caxan] can be analyzed and transcribed *cakan or
*sakàn with the reduction of spirants rule. Kherwarian has the expected sahdn. One could
suggest that the historical development of Kherwarian from North Munda also showed a
reduction of spirants development, but one that worked the other way, i.e., the second,
not the first spirant, was de-spirantized, to h, and in some dialects of Mundari further
reduced (despirantized?) to 9, i.e., saan.

423

With regard to the k >h development in Kherwarian, conditioned (partly) by a glottalized
vowel environment, the original basis for the k > h change in Kherwarian may have been a
reduction (glottalization/aspiration = spirantization) rule, where the X was spirantized in
glottalic environments, and the (spirant) glottalization/aspiration conditioned the loss of
the spirantized X. Thus, Kherwarian and presumably NM *jokaX > jokhaX > joxax >
johaX?” later reduced to joha, ‘cheek’. (The Korku cognate joka has no low tone, and this
would need explanation.) The glottalization itself was then lost in certain environments.

4.2. Creak in KM

The kinds of correspondences between Sora and Gorum, Juray and Gorum and between
SJG and GRG seem, on the surface at least, miscellaneous, but in fact a clear set of features
emerges from these various correspondence sets, namely:

1) correspondences which involve palatality;

2) correspondences with glottalic constriction of some sort (including periodic interrupted
production of glottal ‘stop’);

3) some sort of retraction of the tongue-root;

and possibly,

4) ‘spirantal’ or fricative character.

These points need to be elaborated somewhat. For example, we get SJ *j2/Go. H, SJ *1/
Go. H, SJ *r/Go. H, SI *r/Go.j, SJ *1/Go.j, and even J. rt/Go. Ht. In each case, however
different, palatality appears to be one feature.

In the case of giottalic constriction, we find some sort of glottal (e.g. a stop or H) corre-
sponding to some glottalized consonant. Otherwise, we find glottalized consonants in
correspondence where the point of occlusion does not correspond, as it does in the regular
glottalized consonant correspondences. Or, as occurs in a number of cases in Gorum, while
a ‘regular’ glottal stop, i.e., SJG *-g is associated with automatic nachklang of the vowel, in
these cases, the glottal stop has no associated nachklang and ends the word as would an
ordinary glottalized consonant, e.g., SJ *asu but Go. asu? (i.e., not **asu?u), GR *Vsi, Gt.
asi ‘ache, hurt, be ill’.

As noted earlier, the H (at least in the spectrographic materials we have from Juray) as well
as our phonetic perception of it, is notable for its periodicity, showing up at lower fre-
quencies than normal fricatives, but indicating a clear slowdown of beats of the larynx;
there also seems to be acoustic energy which shows up in the spectrograms as interruptions
or gaps in all the formants. We can consider this spirant-like, since the occurrence of ‘noise’
in the upper frequencies, the periodicity etc. are all reminiscent of acoustic descriptions of
‘fricatives’.

Finally, the effects of ‘creak’ are not limited to a single segment; they affect at least one
morpheme (word?), and may affect other words in the phonological phrase as well. All of
this points to a register-like character. That is, we do not assume KM creak to be a separate
‘phoneme’, but rather we tend towards reconstructing a constellation of features which act
on a word, or minimally on a morpheme, not unlike descriptions of ‘register’ in MK.”

424

4.3. Evidence of Palatality

There is evidence for palatality in the vowel correspondences: Sora /4/ and Go. /e/. Deriving
Gorum /H/ from *j is unconvincing for a number of reasons. First, as noted above, the Sora
CF -9 instead of the expected **-5j presumes an evolution of perhaps *oH-yer/-oH where
H > (°) > ¢ finally, but assimilates to -» before palatals. Second, the vowel correspondences
are all erratic; e.g., SJ *jab-mol/Go. jeH-mol makes no sense in view of forms like S.
jab-lay, Go. jab-lay- (‘stick out the tongue, etc.”) so that it is clearly not the preceding
palatal which produces the otherwise palatal environment required for correspondences of
afe.

The comparative evidence from GRG and from MK suggests that there is no 7’ before e.g.,
m in e.g., jab-mol : jeHmol and in fact there is nothing at all but an ‘automatic’ V (cf. e.g.
Svantesson on Khmu? (1982:97)).

We cannot even, for SJG itself, suggest a reconstructed *j both because of the So. -2, hence
SJG *oH and the fact that /H/ in Go. is idiosyncratic in that environment, since *joj-mol
(or jo?-mol) could conceivably have been the form for ‘seed’ but is not.

Similarly, with SJG *biyoji/Go. biyoHgi, an original SIG */g/ 7 *j/-i in SJ, remaining /g/
in Gorum, so that the original form did not have a palatal in what is masquerading as a
palatal environment.

Possibly Go. H corresponds to an SJ *j of some sort which one probably ought to distin-
guish from ordinary SJ */j/ with Gorum reflexes, hence our hypothetical *j2.

Once we go beyond SJG, there is no evidence of any 'consonant', palatal or otherwise.
However, further indications of a palatality feature and found in other probably correlated
‘palatal’ consonant reflexes elsewhere, e.g., SJ *r : Go. H, SJ *r : Go. j, etc.

Consider a) J. toy-ar- ‘lead, dance, drag by the hand’: Go. aH-toy ‘id.’ with simple SJ *ton-:
Go. toy- Jead’ as well, where toy with retroflex ¢ presumes foy- with a t conditioned by the
preceding H in Go. Retroflex ¢ in Gorum appears to be conditioned by the ‘retraction’ of
the tongue or tongue-root in several other forms in Gorum.??

b) The r : j correspondences are relevant here too since if palatality seems to be intimately
connected with the occurrence of H (SJ *j2) we cannot ignore the areal support for palatal
r as well in Malayalam and Irula. There are r : / correspondences too, however few: witness,
J. lerpoy : Go. lej-puy ‘spare tire, rolls of fat at the waist' or S., J. ler- : Go. lij- ‘spread,
grow big, grow a new sprout (of root vegetables)' hint at a retracted, palatal ofsome sort.
The recurrence ofr :7, r : H as well as assimilated glottalized obstruents with H point to
features of a) palatality and b) glottality.

GRG too, shows palatal consonants where Go. has H, e.g. the previously cited GRG *bier :
SJG *biyaH, and GRG *se2r ‘to sing, shout out, raise the voice’ : S., J. ar-, Go. aH-go-
‘have a hoarse throat, etc and the Gu. irig pre-GRG *Vrij : Go. ariH, arig, ari? ‘millet,
probably Panicum’.

Going outside of SM, there are rare alternations of final r and j in e.g., Ko. namur, namuj
‘to set (of the sun); to sink (drown)’. Note also the presumably doublet forms in Ko. solod,
solor ‘to slip’, where the latter form has an alternate form with a high front vowel, silir.

425

44. Tongue-root Retraction

As to tongue(-root) retraction, as noted above, H in Go. falls into the same category of
conditioning factors as 7, /, (and even g, and k) for the rise of retroflexion of 1.”

While the evidence is hardly overwhelming, it certainly bears consideration in determining
the characterization of the possible phonetic properties of the reconstructed ancestor of
creak. With reference to register, which has been discussed in detail*’, with regard to
tongue-root position we would like to be able to identify ‘chest’ register in MK with NM
low tone/aspiration (i.e. low register?) and perhaps KM V; and to claim that some of the
characteristics Gregerson describes for MK chest register apply here.

According to Gregerson there seems to be a correlation between voiced and lax consonants
on the one hand and voiceless and tense consonants on the other. He claims that advance-
ment or retraction of the tongue-root can constitute a major airstream regulator.

We wish to claim here that whatever the ancestor of ‘creaky voice’ was, it was associated
with the feature ‘retraction of tongue-root’. It seems pertinent to note that many of the
effects associated with such tongue-retraction, namely retroflexion, ‘centralization’ (and/or
‘peripheralization’) of vowels, lenition or weakening of articulation of associated conso-
nants, etc. in a word are also associated with the effects of so-called ‘chest register’.

Cognate reflexes, e.g. Go. /H/ : GRG *r at first seem dubious, but if we hypothesize some-
thing akin to chest register for KM with correlated tongue-root retraction this makes the
correspondences of Go. H : (SJG V) with GRG *r, SJ *r and Go. H (and probably SJ *r
and Go./) more plausible. When the tongue-root is pulled back, the rest of the tongue must
also be retracted. This necessarily includes the tongue blade and dorsal area as well.

4.5. Glottality

The third feature we would like to claim for the historical source of creak is one of glottalic
stricture. Aside from the obvious reflexes of glottal stop, V, and h, in GRG there are
reflexes with pre-glottalized b', d', and g' (this latter before high front vowels derived from
pre-GRG *j).

Presumably, the bilabial closure in e.g. tab-me is in assimilation to the following labial. It
may owe something as well to the preceding vowel. Thus, the glottality is retained and
secondary features of occlusion develop.

While as a first approximation one is tempted to describe the historical source of creak
simply as creak or creaky voice (since these do not otherwise have a place in the recon-
structed phonemic system) this is inadequate to account for the variety of vocalic and con-
sonantal correspondences we have examined here.

There is a polarization and reduction of the three features of the proto-form so that in
cognate pairs while one does not find the entire constellation of features together, one does
find two of the three, usually palatality and glottality (j : H), or retraction and glottality
(r : H), or retraction and palatality (r : 7).

It is not surprising that a complex (and relatively unknown) feature-complex (?phoneme)
should be simplified in a variety of ways.?? However, unlike Proto-Indo-European laryn-
geals, the phonetic nature of which is still much in dispute, the phonetic feature-complex
which we postulate hcre is eminently pronounceable and presumbably reflects a com-

426

paratively recent proto, whereas the Proto-Indo-European laryngeals are more ancient and
more hypothetical.

4.6. Spirantization

Further, and crucial to our hypothesis regarding s-loss as a function of multiple spirant re-
duction, we would also suggest that a feature of ‘spirantization’ or ‘fricativization’ charac-
terizes this entity as well. This feature-complex not only would account for the great
variety of reflexes, but presents no problems in phonetic realization.

It becomes clear that what turns up as creak in certain of the modern languages, low tone
and other vowel qualities in other languages, and a variety of glottalic consonants in several
of the other languages needs to be examined in terms of its possible (i.e., pronounceable)
properties in a reconstructed KM.

4.7. Summary

To sum up, we find a number of associated ‘features’ in most of the cognate sets which in-
dividually seemed unrelated to each other, but which a closer examination showed worthy
of consideration as part of a single entity. Whatever produced creak, register, palatals,
glottalized consonants and A can be shown to be a plausible single phonetic construct not

unlike one or more of the hypothesized IE “laryngeals’.??

That is, it is perfectly possible to produce a sound with the following features:

1. glottalic constriction,
2. tongue-root retraction,
3. palatality,

4. spirantization.

To our knowledge there is no comprehensive treatment of s-loss: where and how it took
place in the numerous languages in which it has developed. If one tries to arrive inductively
at a causal factor or factors for s-loss development in the various independent instances of
loss, one finds no obvious common denominator, not even within a particular language
family. A system with a single spirant seems to be common; in such a pre-s-loss system, in
particular contexts and circumstances, the s may weaken its articulation, and reduce (usual-
ly) to A, which is then itself often lost. In SJG other (e.g. accentual) explanations of
s-loss are inadequate. We postulate the presence (in specified environments) of creak, and
identify its phonemic properties, making use of the GRG forms cognate with those in
Gorum to reconstruct these properties. We claim that s and creak are the members of the
SJG spirant set, and hypothesize a process reducing multiple spirants to one. Thus SJG s in
the (specified) environments where creak is present reduces to @. Phonemic creak itself is
then lost except in some environments in Gorum. This provides an account of one piece of
the phonology of Koraput Munda which makes historical sense of the forms we find in the
data, i.e. in the six contemporary daughter languages.

427

Notes

Munda is spoken exclusively in India, over a large area. There are roughly 12 languages in the family.

Greater Mon-Khmer (which may or may not be divided into two branches coordinate with Munda) is

spoken throughout Southeast Asia from Malaysia through South China, and west into Assam.

We use the following symbols and abbreviations throughout the paper:

SIG Proto-Sora-Juray-Gorum which is the proto reconstructed for Sora (So.), Juray (J.), and
Gorum (Go.), three Koraput Munda languages spoken in Orissa and Andhra Pradesh.

GRG proto for Gutob-Remo-Gta’, also a Koraput Munda sub-group.

KM Koraput Munda is the proto for these six languages

Ko. North Munda language: Korku

Sa. North Munda language: Santali

Mu. North Munda language: Mundari

Ho North Munda language

Khw. Kherwarian is the proto for Sa., Mu. and Ho.

NM North Munda

PM Proto-Munda comprises NM, KM, Ko., and KJ

KJ Kharia-Juang

Kh. Kharia

Ju. Juang

Gu. Gutob

Re. Remo

Gt. Gta?

Other symbols used throughout the paper include:

V which represents an “automatic” (i.e. unspecified) vowel,
L which represents a continuant (e.g., 1, r, n, y, m), and

Ç which represents a glottalized consonant (b’, d’, g’, orj’).

In the modern languages, Juray and Gorum, creak (V) is represented by H.

We deal here only with s-loss in SJG. There are cases of s-loss in Gorum where SJG retains the KM
*s, On these see A. Zide (1982 and 1980). Another, perhaps related, change in SJG consonantism
is KM (and PM) *t > SJG *s (probably before high front vowel(s)). See also Footnote 8.

Pinnow’s remarks on s/@ are, as usual, worth attention (Pinnow (1959), p. 244). Those of his cognates
not listed and discussed here, in our opinion, are not cognate. or do not belong here.

While the SJ cognation is quite clear, the Gorum form presents problems in that the final -d is unex-
plained as a cognate with SJ *?. Although ?/d correspondences are common in GRG this seems to be
an isolated instance in SJG. However, it should be noted that a) various cross-correspondences be-
tween the glottalized consonants -b', d`, g’, j’ with ? in both directions are common in SJG; b) this
“louse” word as -id occurs in Gorum in several forms, e.g. agid “louse”, but also kui-id “louse newly
hatched from egg" and possibly pog(-Jid- “to extinguish a lamp”, which on the surface would not
seem to be cognate, but the act of extinguishing involves pressing the flame (or wick) between the
thumb and forefinger in a motion similar to that used for killing lice. Further, Gta? po-si, presum-
ably from *pon-si- means ‘to crush lice between the thumbnail against the ball of the forefinger’.

It should be noted in this connection, that Munda cognation involves motions and actions rather
than end-results. Cf. as another example, Go. ir-da?- ‘to spin’ with So. gir-da?- ‘to slap the thigh’ and
Gu. girag- ‘to spin’. (Spinning by hand is effected by slapping the fibers against the thigh in a rolling
motion.)

What seems noteworthy about the SJG correspondences is the fact that the Sora CF (the combining
form of a noun attached to verbs) which one would expect to be a regular -2 is, in fact, -9 which
would be possible only if the FF (full form of the noun) had had a glottal stop instead of a j finally.
Since an hypothetical So. *9?-yer is also possible but is not in fact what we find, relegating it to a
simple glottal stop is impossible.

The IA words for ‘winnowing fan’ (perhaps from IE as well, see Mayrhofer 1976:366) Hindi süp,
supli, Oriya supa, Skt. Stirpa (Turner, Item No. 12573, 1966:729) look something like GRG. *-sob,
with preglottalized - 5, but are presumably unrelated. (Kharia supli ‘winnowing fan’ (Bhattacharya
1983: 104) is from Hindi supli.) A Dravidian etymology (Burrow, 1945:118) is so.

428

10
il

There is a question in disyllabic morphemes as to which vowel (or vowels) are to carry the creaky
quality. First, can we say that the morpheme as a whole has creaky vowel quality, or that the vowels
of disyllabic morphemes are independently capable of carrying creak? If the latter is the case, one
would assume that the s-loss would be conditioned by an immediately following creaky vowel. It
seems likely that creak is a property of morphemes (as is e.g., register).

In Gorum, which preserves creaky voice fairly commonly, it is frequently a morpheme itself (e.g. of
affectedness in verbs and/or a junctural phenomenon). From the KM point of view, it seems plausible
that creak could be a feature of morphemes in which case it could be written once on the final
stressed vowel.
SJ *AtH, Re. suti?, 'female genitalia', does not meet the CVL requirement. There is at least one other
reason to find the cognation questionable: there is a remote possibility it could be a borrowing in
Remo (from Desia c/sotli) but this is by no means certain either. Also, its morphemic analysis is
unclear. It is not certain whether it is one morpheme or two. If it is two, the second would be -ti?.
If the cognation were incontrovertibly demonstrable this would require changes, e.g., of morpho-
phonemic constraints on canonical forms conditioning historical rules, but would not thereby un-
dermine the creak hypothesis.
Contrast Go. sam, Re. sam ‘speech, word, utterance’, Gu. samo, Gt. samwa ‘language’.
NM *sVreXfi or *sVriXfi where X indicates low tone, aspiration, or (low) register.
It looks as though there could be a residue of a formerly possible phonemic H even in Juray. Sora
has no evidence of creak, however, phonetically or phonemically.
The symbol j2 is used to indicate correspondences between glottalized consonants in SJ which have
assimilated to the following consonant (frequently but not exclusively to -b’ before a following -m-
initiated morpheme) but which display vowel correspondences preceding the glottalized consonant
which are otherwise exclusively found in palatal environments. SJ *a/Go. e correspondences are
understood as SJG *e/-palatals. SJ */2 almost invariably corresponds to Go. H, i.e., does not cor-
respond with Go. j. ‘New’ and ‘paddy seed’ above are good examples.
There is an SJG *gud- ‘to scratch’. As in most Munda languages, a CVC verb may be reduplicated to
indicate a) repeated action and/or b) intensity of action, so that gugud- and gudgud- (R-gud-) are
regular forms in e.g., Gorum where they mean either ‘scratch over and over again’ or ‘scratch hard’.
*guHsgud- in Gorum could be understood to come from a reduplicated gud- via gu ’gud with a ‘weak-
ening’ of glottal stop to H. However, the form we in fact find for ‘to scratch oneself’, a reflexive (i.e.
not an intensive) is guguH- (not **guHgud-). The problems this presents are that a) there is in fact no
d in the form to dissimilate from, the final ‘consonant’ is H, not d; b) elsewhere, H is a morpheme of
affectedness of subject and the reflexive meaning certainly can be seen to come under such a semantic
rubric; and c) the ‘infinitive’ form is guguH-nuH, unusual in its permissible two glottalics within a
word (the ordinary situation is for one of the glottalics to disappear); d) the correspondences for
SIG are equally irregular: Ju. gu gud- is by no means an ordinary reduction of d. Further, the use of
-dam, the regular and commonly found reflexive verbal morpheme is not evident. Sora ged-, as well,
it totally unexpected. Jt has had a shift of vowel in a language otherwise not known for any instances
of ‘ablaut’.
What seems to be one explanation is that there is probably a reflexive (or affectedness?) infixed
morpheme which has left decided traces on the internal makeup of the verb, and that either we have
an instance of two distinct and separate verbs, one with the meaning ‘to scratch’, *gud- and a second
verb with the meaning ‘scratch oneself’ ???*R--goH, or R-gVH- (SJG *o >S. e, J. o, Go. u). What
seems far more plausible is that we have a verb *gud- ‘to scratch’ with an infixed morpheme of sub-
ject-affectedness (which exists in modern Gorum), and *H as a register-like morpheme on the word
which affects a) vowel quality, hence So. e, and b) other glottalic consonants (e.g., -d’) within a
word.
This, in contradistinction to Wang’s [h] which he calls a voiced glottal fricative (1968: 6) but which
is really an approximant (cf. Ladefoged (1967:88 ff.)). Gorum [H] is much more highly constricted
and thus closer to actual glottal closure (or glottal stop) with which it, on occasion, in certain
specifiable environments, alternates. Ladefoged (1964:16) describes laryngealized voice (Catford’s
‘creaky voice’) in the following way’

“In this state of the glottis there is a great deal of tension in the intrinsic laryngeal muscul-

ture, and the vocal cords no longer vibrate as a whole. The ligamental and carytenoid parts of

429

the vocal cords vibrate separately sometimes almost exactly 180° out of phase with one an-
other, one end opening as the other is closing. This produces an apparent increase often an
approximate doubling, of the rate of occurrence of glottal pulses. ... Laryngealized voicing
need not occur in implosive consonants, and equally it can occur without the downward
movement of the larynx which must (by definition) be present in an implosive. We can there-
fore, separate out two kinds of glottalized consonants: implosives ... and ... laryngealized
consonants (as in Hausa). ... It is often not possible to make an absolute distinction between
laryngealization and glottal closure; as in the case of so many other phonetic oppositions
there is an infinite gradation between the poles of the two categories. .
Recordings and spectrograms were made of select sets of minimal pairs of vowels and of the glottal-
ized consonants, glottal stop, and creaky voice in the speech of D. Gomango and B. Boya, Juray in-
formants, at the Central Institute of Indian Languages, Mysore in February, 1980. Thanks are due to
Dr. D.P. Pattanayak, Director for making available to A. Zide the excellent facilities of the CIIL.
Thanks are also due to Nagarajarao, Jennifer Bayler, and B. Rajpurohit for their help with the
technical work, especially for the preparation of tapes, and spectrographic materials and for the
analysis of the spectrograms of the vowels and of creaky voice.
The alternation of & with ? in Remo also occurs for forms where glottalization was hypothesized,
*gel, geXl- ‘to make offerings, do puja to a god’. *sol also had been reconstructed with glottalization
(X) by N. Zide (1965). What relationship obtains between glottalization and creak, and more partic-
ularly, how we can reconstruct *sol- and *geXI are relevant questions we cannot now answer.
With regard to bier in Gutob, see Ramamurti’s Gutob form bar-tog (Ramamurti, 1938), and N.
Zide’s comments (1978) made with no thought of bier and its relevance.
In this connection, cf. too So. yer-, J. jf- ‘to go’ which become more plausible as possible cognates
for Go. uiH- especially in view of the r/H correspondence sets we discuss later in the paper. (But note
Gt. wir ‘to run.)
Note that the preceding syllable (morpheme) in GRG *kV-sob (KM *sə) ‘winnowing fan’ exhibits
strange vowel correspondences. Gu. ki-sob, Re. ke-sub, Gt. (Plains) ki-so?, Gt. (River) ké-so?.
For GRG *tVme and *cVmol have a regular automatic vowel in the first syllable, which has creaky
voice as has the whole morpheme. The vocalism in SJG is something else. It is not clear whether the
GRG automatic vowel is to be reconstructed for KM, or something closer to the SJG *VH.
Here too the vocalism, i.e., the correspondences between a/oH and a/eH are unusual, the normal
correspondences between SJ o and Gorum being 2/0.
Cf. A. Zide, S-loss in Sora-Gorum (in press).
How one treats glottalization of vowels is not unrelated to this discussion and something will be said
about this further on. /h/ is probably to be reconstructed in certain expressive functions, but not asa
regular phoneme of any widespread distribution. ^ does occur, but as a reflex of PM *k in Kher-
warian and in Gta? and possibly in pre-Gorum in restricted environments. One does find [x] in Juray
and Korku, but they are clearly instances of restricted allophones of k and Kh respectively in true
medial position. There is no question of their being independent phonemes.
Note too Go. aH-si-nuH ‘wash the hands’, aH-jig-nuH ‘wash the feet’ and aH-mad-nuH ‘wash the
face’ in connection with Ko. ab-.
The situation in the Indo-European languages (much of this we owe to Eric Hamp, personal com-
munication, 1983) seems quite different, and suggests no solution to the Munda s-loss problem. In
IE, for instance, there is no vowel breaking associated with s-loss (in the IE families which undergo
$doss), nor any multiple spirant conditioning. There is, however, in both PIE and PM a similar pre-s-
loss 'system', i.e. one with a single spirant, and laryngeal features (for Munda one, or possible two,
we will not invoive ourselves in controversies about how many, which, etc. IE laryngeals there are).
And these ‘laryngeals’ are similar in both families in that they account for odd vowel correspondences
(affect vowel coloration, and raising and lowering), aspiration and glottalization. In Gorum there is
something paralleling the development from Iranian -VsV- to Avestan -yh- ([fi]?). In Gorum, in a
small number of forms, A. Zide reconstructs -h- for pre-Gorum in order to account for both mor-
pheme-initial earlier k and s which show up in Gorum as -y-. Thus, Go. ayon, ‘son, child’, presumably
from pre-Go. *Vn-hon, So. 272, Ju. a-on, NM *kon, Khw. *hon; Gorum ayid /-id, but SJ i7i/-i. The
Gorum form is presumably from pre-Go. *Vn-siH(?). We find in Gta? h from earlier y, thus GRG
*tV yon, Gt. thwä ‘to stand’. (The SIG form metathesises the z and the y, i.e. SIG *tVnan.)

430

% In contrast to voiceless aspirates, voiced aspirates in parallel contexts (e.g. ghal- ‘to show’) behave

differently: by a general rule in Korku no voiced aspirates occur non-initially, and the reduplicated

form of ghal- would be gha-gal, the aspiration of the second g being absent, but the low tone (of the

low tone/aspiration phoneme) being maintained. The low tone automatically occurs with medial

aspiration, thus kakhad above is phonetically [kaxad]. See N. Zide (1960) for further discussion of

interpretations of "low tone/aspiration’ in Korku.

Here X is meant to represent glottalization.

Treatment in terms of prosodies, long-components, etc. could be provided. For descriptions of

Since a preceding /r/ exists as one of the conditioning factors for the rise of /t/ in Gorum, this makes

it unlikely that J. ar-, Go. aH- are to be rejected as distinct affixes.

Note that Gta? (but not GR) also has a retroflex 7 the origins of which are obscure, but that retro-

flexion seems to be correlated with “glottalized” vowels and perhaps with pre-glottalized b in partic-

ular. Note too GR *bVton- “be afraid", Plains Gt. bro, River Gt. btfy (SJG bVton-). The *o in GR

was interpreted (N. Zide, 1965) as a glottalized vowel.

See Gregerson, 1976, and Shorto, 1966 where, in reference to Mon, he says,
“The paratonal register distinction is broadly similar to that described for Cambodian by
Henderson. Its exponents are distributed throughout the articulatory complex but exclude
pitch features. Chest register, symbolized by a grave accent placed over the vowel (ket, hakoa)
is characterized by breathy voice quality in association with a general laxness of the speech
organs and a relatively centralized articulation of vowels. The more frequent head register is
unmarked in the transcription (ket, hakoa); it is characterized by a clear voice quality, relative
tenseness, and peripheral vowel articulation.”

Note the fate, for instance, of Indo-European labiovelars in the various Indo-European languages.

The Munda features in any case are much closer to the spoken languages than are the IE constructs.

One wonders if anyone anywhere has suggested that PIE may be like some of the MK languages, i.e.,

a ‘register’ language.

Instances of such loss of s via A are evident in Caribbean Spanish and Andalusian Spanish. In Puerto

Rican Spanish the change from h > $ has taken place in the last twenty years. Cf. also note 25 above.

The s >h change is a common one. In the Indian area we have cases from Rajasthani dialects, Bhili
dialects, Bengali dialects, Assamese dialects and Sinhalese (Sinhala) as well as from Central Dravidian.
Only the last might be relevant to a (partial) historical explanation of the SIG sound change. Note
that the Koraput area (and neighboring regions to the south) lacks aspiration: the Indo-Aryan Desia
(Kotia Oriya’) lacks it, as do five of the six KM languages (the exception is Gta? where A reflects $M
*k as well as SM ‘expressive’ 4), and the Dravidian languages of the area, including non-standard
Telugu.

References

Bhattacharya, S., 1983. "Excerpts from the Kharia Field Note Books", International Journal of Dravi-
dian Linguistics 12.

Burrow, T., 1945. TPS (cited thus by Mayrhofer).

Catford, J.C., 1968. The Articulatory Possibilities of Man. Manual of Phonetics (Amsterdam: North-
Holland Publishing Comp.).

Gregerson, Kenneth J., 1976. “Tongue-Root and Register in Mon-Khmer", Austroasiatic Studies Part 1
(Honolulu: University Press of Hawaii).

Ladefoged, Peter, 1967. Linguistic Phonetics (Los Angeles: UCLA Phonetics Laboratory).

Mayrhofer, Manfred, 1976. Kurzgefafites Etymologisches Wórterbuch des Altindischen, Vol. 3 (Heidel-
berg: Winter).

Pinnow, Heinz-Jürgen, 1959. Versuch einer Historischen Lautlehre der Kharia-Sprache (Wiesbaden:
Harrassowitz).

Ramamurti, G.V., 1938. Sora-English Dictionary (Madras).

Shorto, H.L., 1966, “Mon Vowel Systems: a problem in phonological statement”, In Memory of J.R.
Firth (London: Longmans), 398—409.

431

Svantesson, Jan-Olof, 1983. Kammu Phonology and Morphology (Lund: Liber Fórlag).

Turner, R.L., 1966. A Comparative Dictionary of the Indo-Aryan Languages (London: Oxford Univer-
sity Press).

Wang, William S.-Y., 1968. The Basis of Speech (Berkeley: University of California Press).

Zide, Arlene R.K., 1980. Revision of a paper presented at the 1978 Mysore Second International
Austroasiatic Congress. See A. Zide in press.

Zide, Arlene R.K., 1982. “When a glottal stop is not a glottal stop”, paper presented at the Chicago
Munda Languages Conference, May 1982.

Zide, Arlene R.K., In press. **S-loss in Sora-Gorum", Proceedings of the Second International Congress
of Austroasiatic Linguistics (Mysore: Central Institute of Indian Languages).

Zide, Norman, 1960. Korku Phonology and Morphophonemics, unpublished doctoral dissertation, Uni-
versity of Pennsylvania.

Zide, Norman, 1965. ‘“Gutob-Remo Vocalism and Glottalized Vowels”, Lingua 14:

Zide, Norman, 1978. Studies in the Munda Numerals (Mysore: Central Institute of Indian Languages).

Olr. ‘tab(a)ir ‘brings’, ‘tait ‘comes’

Eric P. Hamp (University of Chicago)

1. The first vowel of 'tabfa)ir, verbal noun tabafi)rt, has been the subject of notice by
careful accounts of Old Irish; see GO/ 53, $ 82. It has been supposed that the a spread
from the perfect ‘tarat ‘has given’ < to r[o)-ad-d- (cf. Celtica xi, 74) < (to+)ro-ad-da- <
*pro{ad-)dö- < deH,- ‘gave’ (cf. Studia Celtica xiv-xv, 1979-80, 107). A similar spread
could equally be justified, but on slightly different semantic/syntactic grounds, by starting
from *pro-ad-dha- < *ad-dhHe-. In this case I would equate *ad-dha- with Phrygian aödaker,
and explicitly because I consider Phrygian genetically a Western IE language; I refer to the
well known clause exemplified by ws ra uarka Kakovv addaker (Ti) and oc D oeuor
kvouuave, kakov a[5]6aker. It is possible naturally to link the sense ‘puts, places, sets’
(DIL D [1959] 208—10) with a reconstruction *pro-ad-dha-, but it must also be granted
that semantically this sense could readily be derived from deH,-. Yet we shall see that there
is considerable argument for both these pre-forms.

It has been pointed out (GOI § 82) that the archaic pret. tubert ‘tubart does not show a,
with the implication that the spread from ‘tarat had not then taken place. But at least one
of these (contubart Thes. II 242,20 = Lib. Ardm., even though *e is not preserved here)
clearly means ‘gave’. This form therefore does not, strictly, tell us anything about the
prehistory of the value "tab(aJir ‘brings’.

It may be suggestive of the background of 'zabfa)ir ‘brings’ to recall that in Later Phrygian
aßßeper and affeperop were apparently functional equivalents of aödaker and adéaxerop.
Now we know that the pre-form of tabhair cannot have been simply *ad-bhere > *(t(0)+)
abbere; the phonetics will of course not support this. Moreover, I have shown (ZCP 39,
1982, 209 and 212—14) that *ad-bher- in Celtic meant ‘offer’. Nevertheless, the vocalism
of *t(o)-a-ber- suggests that at an earlier time the to, which we know not to have been
originally a preverb, may have taken the place of an older moribund preverb. The further
thought suggests itself that a transformed match for *#/o)-a-ber- > *t[o)-ad-br- may be
found in the suppletive Welsh imperative dabre ‘come!’; cf. Lat. te adfers, urbem adferimur.

I therefore suggest the equations

Olr. ‘tarat < *pro{ad-)d6- = Welsh rhoddi = Latin prodo = npoëiô cou

Olr. ‘tarat < *(pro)-ad-dha- = Phryg. addaxer = Latin addö

Then the aoristic *ad-dheH,- would have found itself in a natural suppletion with a com-
pound of the presential *bher-.? Thus we may equate

Olr. ‘tab(a)ir : Welsh dabre : Phryg. aßßeper : Lat. adferö.

Underneath this last equation I propose that there was an old compound (i.e. complex
verb, or collocation) antedating the productivity of the Western IE *ad, with a particle

434

TOWARDS ending in a syllabic and which could have produced Celtic *[a]. We find the best
match for this in Iranian: Sogdian "Br- ‘bring’, "B’yr 3sg. impf. < *a-brya- ‘was brought’,
suppletive to "at, e *a-gata- ‘come (used as a transitive)’ (I. Gershevitch, A Grammar of
Manichean Sogdian, 1954, 88 540, 603, 653)? This selfsame suppletion is continued in
modern Yaghnobi wr- ~uxt-. The compound occurs with clarity in Avestan a-bar- and in
the Pamir dialects e.g. Yidgha — Munji avar- ‘bring’.* The exact vocalic background of this
Indo-Iranian d is agreed to be obscure. If in origin the particle was "(HJeH, a zero-grade
would have produced *2 > Celtic *a. Such a slender body would account for its early
demise.

We recover, then, *pro + deHo- ‘give away’, *(H)eH + dheH,- (> *ad + dheH,) ‘place some-
where’, and */H)JeH + bher- ‘bring’. Apart from the usual appeal to (morpho)phonology,
syntax and semantics, we reconstruct these preverb collocations by exploiting criteria of
morphological suppletion and Indo-European dialectology.

2. Indo-European had an old pattern of expressing the notion ‘come’ by a collocation
TOWARDS + GO. Thus Polish iść ‘go’: przyjsc, dojs¢ ‘come’; Serbo-Croatian ici idem ‘go’,
otici ötidem or Odem ‘go away’, dôci dôdèm ‘come’. Similarly, Old Irish matched téit
‘goes’, tiagu ‘I go’, subj. téis, fut. rega, lotar ‘they went’, perf. do‘coid ‘dechuid etc. with
doter ‘comes’, do tiag “I come’, subj. do tias, fut. do ’rega, do lotar ‘they came’, perf.
do dechud etc. Notice that we appear to have do « to * GO. The prototonic forms of
‘come’, however, show ‘fait ‘comes’, ‘taig ‘taeg ‘I come’, subj. ‘tdes, fut. ‘terga tirga,
‘tultatar ‘tullatar ‘they came’, perf. ‘tuidched etc.

The consonantism of prototonic ‘come’ is explained by a rule (GOI § 179) applying to
sequences C, VC, ..., but Thurneysen is obliged to reconstruct (schematically) *ta-thet-
etc. < *to-. The a is attributed, improbably, to the impv. 2sg. fair, which in turn is sup-
posed (GOI § 588) to be from t-air-ic somewhat anomalously. For ‘tair as a shortened form
of the 3sg. pres. subjunctive of do airic ‘comes’ see V. Hull, Lg. 25, 1949, 132. If we can
explain ‘tait, 1 propose that likewise fair ‘come!’ is to be given a simple relation to eirg/g)
‘go!’, *t(o)-a-erge. Note that the outcome calls for *['a], and cf. below.

Observe now that ‘ta- appears only in the presential system, largely */s)teigh-, of ‘come’.
This agrees well with ‘tab(a)ir, and we know that semantically BRING/TAKE ~SEND may
be viewed as factitives of COME/GO.? Moreover, the perfect *di * ko-ued- > ‘dechuid
includes preverbs which could have rejected an old *-a-.

Analogously to */H)eH + bher-, therefore, I compare Olr. ‘tait ‘comes’ to Skt. a + gam-;
‘taeg ‘I come” is therefore *#{o)-a-teégu, with *t(o)-'a-.° That is, if SEND = GO-FACTIT and
COME = TOWARDS + GO, then BRING must be COME-FACIT = TOWARDS + GO-
FACTIT. Thus we may quite simply derive
TOWARDS + GO > (to)-a-tég-;
TOWARDS + GO-FACTIT >

( to)-a-ber-
It is expectable, then, that the suppletion in these two verbs should be homomorphous or
coextensive; for at bottom it was identical. And the ancient preverb may be said to have
participated here in but a single abstract collocation. The fundamental correspondence is
therefore that with Indo-Iranian @ + gam- (and the suppletions of *gWern-). In view of the
presumption of the presence of *H)JeH > *a- and of its insubstantial moribund status, I

435

submit that we may now understand how the stem of Skt. gam- ‘go, move, depart’ came to
mean ‘come’ in Latin wenid and the Germanic etymon which come represents. That is, the
simple base of IE *gWem- ‘go’ appears in Western IE dialects in the meaning ‘come’ by
virtue of a loss of the once semantically essential and distinctive *(H)H > *a probably
through sandhi phonetics or ambiguity. In Italic and Germanic this slender particle/preverb
was not protected by a sentence connective as we see that it was in Celtic.

Notes

The most direct derivation of this compound is surely from *deH,-, pace GOI 35, § 50 (note), whose
paradigm would have included *dH,- = da-, which would of course have been homophonous with
*dhHe- = dh?- in Celtic. This is justified by the pervasive semantics (see DIL D [1959] 204—8) as
well as by its syntactic behaviour and the separate syntax of *kor- ker-d(h)- (cu(ijrethar fo ceird; cf.
EC 18, 1981, 113 and MSS 40, 1981,43—4), which seems to be the main continuant of IE *dheH,.
On the suppletive nature of *bher-, see ZCP 39, 1982, 205-18. Remember that *pro could be
simply perfective in Celtic.

Further on the idiosyncracies of ‘come’, ‘bring’, ‘give’ see my paper on Welsh moes ‘give me’, Poznan
Conference on Historical Linguistics, 1983, which will be incorporated in a separate publication.

A similar match is found in turn for *ad-dheHe in Skt. d + dhd- ‘place, set, lay, put’. Note the entire-
ly different semantic relation observed in the directionality of Skt. à + dä- ‘take, receive, deprive’.
Yet pra + dā- ‘give up, deliver, bestow’ exactly corresponds to Welsh rhoddi 7 Lat. pródó. These
Indic facts precisely confirm the analyses "(proJ-ad-dho- and *pro-(ad-)dò- above.

Cf. the Sogdian and Yaghnobi above; and Robert Binnick, “On the nature of the "lexical item’,”
Papers from the Fourth Regional Meeting, Chicago Linguistic Society 1968, 1—13.

I elaborate this accentuation with *to, and its implications, elsewhere.

Internet Archive Audio

Featured

Top

Images

Featured

Top

Software

Featured

Top

Texts

Featured

Top

Video

Featured

Top

Mobile Apps

Browser Extensions

Archive-It Subscription

Save Page Now

Full text of "Cardona & Zide (eds.) - Festschrift for Henry Hoenigswald (1987)"

See other formats